Skip to main content

Full text of "Models for Medicare payment system reform based on group-specific volume performance standards (GVPS)"

See other formats


Models for Medicare Payment System Reform Based on 
Group-Specific Volume Performance Standards (GVPS) 



Christopher P. Tompkins, Ph.D. 
Stanley S. Wallack, Ph.D. 
Jon A. ChiUngerian, Ph.D. 
Sarita Bhalotra, M.B., B.S. 
Mitchell P.V. Glavin, M.Sc, M.A. 
Grant A. Ritter, Ph.D. 
Dominic Hodgkin, Ph.D. 

^^^^^^ for 

Heller Graduate School 
Brandeis University 



Final Report 

Submitted to Teresa DeCaro, Project Officer 
Health Care Financing Administration 



June 9, .^95 



This final report was submitted to HCFA under Cooperative Agreement #1 7-C-90129/1-01 . The 
views and opinions expressed in this report are the authors; and no endorsement by HCFA or 
DHHS is intended or should be inferred. 



Table of Contents 

L Purpose 1 

n. Overview of the Project and Report 2 

A. The Project 2 

B. The Report 3 

III. Policy Context and Alternatives 4 

A. State Level MVPS • 5 

B. HMO Risk Contracting • • • 6 

C. Penalize High Cost Hospital Medical Staffs o 

IV. Rationale and Goals for GVPS 7 

A. Rationale 7 

B. Goals for HCFA and Providers 8 

V. Summary of Basic GVPS Models 9 

A. The Policy Context: MVPS 9 

B. Methods to Monitor Volume Performance 10 

C. Rewards and Penalties 12 

VI. Utilization Measures 1^ 

A. The Basic Measure: RPUPS 15 

B. Scope of Service 15 

C. Illustrative Findings 16 

VII. Setting Performance Standards or Targets 19 

A. Scope of Services 19 

B. Differential Growth Rates Based on Location 22 

C. Provider Level Versus Market Level Changes ^ 25 

D. Cumulative Versus Year-to-Year Targets 29 

VIII. Rewards and Penalties 35 

A. Policy Questions and Objectives 35 

B. Options for Rewards and Penalties 35 

C. Reward and Penalty Formulas 38 

D. Transaction Mechanisms 42 

E. Possible Implications 44 

IX. Eligibility Criteria for GVPS 45 

A. Linking to Purposes - 45 

B. Basic Eligibility Criteria 46 

C. Criteria for Managed Care and Preferred Provider Initiatives 48 



X. Provider Responses to GVPS 51 

A. Deciding Whether to Elect GVPS 52 

1 . Meeting EligibiHty Criteria 52 

2. Service Categories for RPUPS 53 

3. Reward and Penalty S' .e 53 

B. Action Plan 53 

1 . Increasing Market Share 59 

2. Managing Cost and Utilization 60 

a. Within the Organization 60 

b. Outside the Organization 61 

3. Care Management Strategies 62 

C. Managing Performance 63 

1 . Management Control 63 

a. Practice Guidelines 63 

b. Feedback Mechanisms 63 

c. Compensation Systems 64 

2. Evaluating Performance 64 

a. RPUPS 64 

b. Patient Capture Ratio 64 

c. Services Provided 65 

XI. Simulations of Medicare Expenditures and Rewards to Groups Under GVPS ... 65 

A. Analytic Framework 65 

1 . Defining Financial Impact 65 

2. Time Horizon 66 

3. Relevant Services 66 

4. Defining a Base Case 66 

B. Assumptions 66 

1 . Assumptions About the Medicare Environment 66 

2. Assumptions about the GVPS Policy Design < 68 

3. Assumptions About Providers Under a GVPS Program 69 

4. Summary: The Base Case Scenario 71 

C. Findings 72 

1. Results 72 

2. Sensitivity to Policy Parameters 75 

3. Sensitivity to Other Assumptions 75 

D. Discussion of Simulation Results 78 

XII. Conclusions and Recommendations 79 

82 

References 

Appendix A: Consultations with Multispecialty Groups A-1 

Appendix B: Data ^"^ 

Appendix C: Alternative Model Specifications C:-l 



i 



Executive Summary 

1. Purpose 

This report discusses potential policy options for Medicare based on Group-Specific 
Volume Performance Standards (GVPS). We developed a basic model and analyzed several 
variations with the following fundamental objectives: 

► To study alternatives to the current system of Medicare Volume Performance Standards 
(MVPS) that would allow qualified physician groups to elect separate performance 
standards; and more generally, 

► To explore ways for HCFA to control the rate of increase in the volume and intensity of 
services delivered to Medicare beneficiaries. 

The proposed models could give providers financial incentives to manage the services delivered 
to their Medicare patients, and reduce Medicare spending for all covered services, not just 
physician and supplier services. 

Under MVPS, fee updates for all physicians can be reduced if annual volume 
performance standards (national expenditure targets for physician and supplier services) are not 
met. There are at least three problems with the mandatory national groupings: 

► Physicians have weak economic incentives to be efficient because individual performance 
is aggregated with the rest of the nation. 

► If national standards are exceeded, physicians who are relatively efficient get penalized 
as //their own performance were average. 

► Over many years, shrinking Medicare payment levels relative to other payers may 
threaten beneficiaries' access to physician services. 

When MVPS was enacted. Congress acknowledged that refinements to the basic 
approach could be warranted, and specifically called for development of group-specific 
performance standards. This report discusses models that would allow HCFA to monitor and 
reward physician organizations separately from the national experience. Furthermore, incentives 
can be expanded to include efficient management of all Medicare-covered services. Given its 
general applicability, this approach may have advantages over other policies such as capitation. 
Whereas Medicare has not achieved savings from HMO enrollment because of favorable 
selection. Medicare cannot lose money on GVPS but would share savings generated by efficient 
providers. In addition, GVPS would encourage the most efficient physician groups to serve the 
most expensive Medicare beneficiaries. 



ii 



2. Goals for GVPS 

An option for a group to elect separate performance standards presumably would require 
a sufficient overlap between the interests and goals of Medicare and "qualified" physician 
groups. This overlap is likely to be greatest with physician organizations that can be relatively 
efficient and that provide high quality care. 

Specifically, we see the following goals for Medicare and GVPS: 

*■ Achieve and maintain high quality health care. Attempts to cut costs can run a --^k of 
compromising quality and access. Medicare beneficiaries should receive services that 
equal or exceed prevailing quality standards in the local community. 

► Improve clinical efficiency. For each beneficiary, the combination of services results in 
the lowest cost necessary to meet the quality standards. This means avoidance of 
unnecessary referrals, tests, and services. 

► Improve market efficiency. This is multifaceted: 

• Relatively efficient providers are rewarded and increase their market share; 

• Inefficient providers bear the brunt of fee reductions or other penalties; 

• Providers generally become more efficient in response to the market dynamics, or 
lose market share. 

► Achieve budget neutrality. Implementation of GVPS does not lead to higher aggregate 
Medicare expenditure levels. 

Providers electing to participate under GVPS may not all have the same objectives or 
priorities. However, participation might help them achieve particular goals: 

► Compensation for clinical efficiency. Savings attributable to clinical efficiency could 
be shared with participating providers. 

► Greater market share. This may be achieved by increasing the number of patients 
served and/or the scope of services provided. 

► Support for new strategic orientations. Market pressures for efficiency create new 
challenges and new opportunities for providers. GVPS could promote strategic 
orientations that include, but go beyond Medicare. 



iii 



In summary, GVPS could enhance equity in payment and create incentives for groups to 
manage the care of their patients. Many providers are seeking ways to defend or increase their 
market share by efficiently managing the volume and intensity of services. GVPS could 
reinforce these initiatives and encourage similar efficiencies for Medicare patients. Efficient 
providers that meet target growth rates would be rewarded. Meanwhile, the burden of financial 
penalties would be more concentrated on physicians who have not demonstrated acceptable rates 
of increase. 

3. Basic Methodological Approach / 

The empirical work in this study is based on two sets of providers. First, we worked with 
a dozen physician groups located in different parts of the country. Physicians and other 
managers at the groups have contributed to the development of workable models for GVPS. In 
addition, we gathered Medicare claims data for patients seen by these groups in order to analyze 
resource consumption at the provider level. Groups were defined at the corporate level using the 
Provider Tax Number, which encompasses all relevant physicians and their billing numbers. 

A second set of 66 providers was selected from the three-digit ZIP Code areas that 
account for at least 5 percent of Medicare patients seen by each of the original groups. In all, 78 
providers were included in the analyses. For each provider, we obtained all National Claims 
History file records for 1991, 1992 and 1993 for all beneficiaries seen during that period. The 
average number of beneficiaries seen in 1992 by the 78 providers was about 15,000. In addition, 
we obtained similar data for random beneficiary samples drawn from the same market areas, in 
order to determine market level reimbursement rates and trends over time. Each provider 
typically saw only a small fraction of all users in the entire market area (i.e., less than 10 
percent). 

From these data, we were able to measure average utilization for all Medicare patients 
seen by a provider during each of the observation years. Utilization was measured as 
reimbursements per patient. Changes in this measure at the provider level were compared to 
changes at the market level. Comparisons were made with and without adjustments for changes 
in health status. 

4. GVPS Models 

GVPS models would supplement and refine the national MVPS, which focuses on rates 
of increase rather than absolute expenditure levels. This is consistent with the policy goal of 
achieving sustainable growth rates in Medicare spending. Proposed GVPS models emulate the 
focus on growth rates, which could encourage participation by providers with relatively sick 
patients and/or elaborate practice styles. 

GVPS models must depart from some aspects of the current MVPS approach. First, 
national expenditures are measured in total, with adjustments for changes in the Medicare 



iv 



population size. At the provider level, it is more valid and convenient to measure performance 
on average for patients seen. Second, current performance standards are set nationally and 
reflect average performance not only across providers but also across market areas. Valid 
performance standards for providers need to reflect local market conditions, not expectations 
averaged across all market areas in the country. Thirc, .... case mix of patients seen by a 
provider can change over time and performance standards should account for such changes. 
HCFA has experience with these issues through its capitation payment system, and we borrow 
■ from some methods already used by HCFA. Finally, GVPS could encourage physicians to 
manage all Medicare services for their patients. Thus, the policy coyld embody a vision for 
physicians to manage their patients, not just their own practices or only professional and supplier 
services. 

a. The Measure of Utilization: RPUPS 

We measure average resource consumption at the provider level by Reimbursements Per 
Unique Patient Seen (RPUPS). In the denominator of this ratio are the Medicare beneficiaries 
who receive physician services from the group during the calendar year. Although beneficiaries 
may be informed about GVPS, we presume there would not be an enrollment requirement. In the 
numerator are Medicare reimbursements to all providers seen by these beneficiaries during the 
calendar year, for the services included in the measure. Two scopes of service are contemplated 
for the GVPS option: those services currently under MVPS, and all Medicare-covered services. 
In either case, providers are reimbursed for services according to all of the applicable prevailing 
payment policies, including the Medicare Fee Schedule for physician services, the hospital 
Prospective Payment System, etc. 

The values of RPUPS range widely among providers. For the 78 providers in this 
analysis, the proportion of all Medicare reimbursements in 1992 that were for physician and 
supplier services averaged 25 percent of the total for all Medicare services used by their patients, 
and ranged from 15 to 39 percent. For patients seen by a particular provider, the share of all 
reimbursements for physician and supplier services that went to that provider averaged 1 8 
percent, and ranged from 2.9 percent to 52.9 percent. We call this the Patient Capture Ratio 
(PCR). •• 

b. The Performance Standards 

GVPS would compare resource consumption in a given year to a target level or 
performance standard that is derived from resource consumption levels (i.e., RPUPS) for a base 
year, multiplied by a specified percentage rate of increase. Targets could be set on a year-to-year 
basis, using the most recent observed RPUPS as a base. However, this has two disadvantages. 
First, groups would have difficulty realizing substantial savings if performance standards were 
"ratcheted down" each year on the basis of earlier successes. Second, groups may have 
incentives to generate excessive utilization in some years alternating with years of lower 
utilization and apparent but artificial "savings." Preferably, targets could be updated 



V 



cumulatively from the level of RPUPS observed in a specified base year, without regard to 
intermediate values of RPUPS. This would allow groups to benefit from implementing systems 
that continue to keep utilization rates lower, similar to Medicare capitation payments to HMOs. 
We explore methods to adjust for differences in a provider's case mix between the base year and 
performance year. 

Rates of growth varied among market areas. Between 1991 and 1993, rates of growth in 
total Medicare reimbursements per patient (unadjusted for case mix) ranged from 6 percent to 24 
percent. For physician and supplier services, rates of growth ranged from minus 1 .2 percent to 8 
percent. Differences across markets are an issue confronted by HCFA in setting payment rates 
for all providers, including HMO risk contractors. We propose methods similar to HCFA's 
geographic adjustment for capitation payment rates, which are based on the beneficiary's county 
of residence. 

Figure ES-1 shows how rates of growth at the provider level compare to rates at the 
respective market level. The origin of the axes represents a perfect prediction at the provider 
level, based on changes at the market level. The horizontal axis shows deviations from the target 
when RPUPS is defined to include physician and supplier services only; the vertical axis shows 
deviations from the target when RPUPS is defined to include all Medicare services. Thus, points 
that are toward the left, and toward the bottom of the figure, are providers with lower rates of 
increase. 

We interpret the substantial clustering of points within 10 percent of the market 
prediction to indicate a strong tendency for RPUPS to be stable at the provider level. For the 
most part, points outside the cluster (such as Points A and B) are providers that experienced very 
large changes in the number of beneficiaries seen during these years (e.g., a 50 percent change). 
We regard these instances as neither data nor policy problems. Rather, they indicate that groups 
will need to inform HCFA about changes in their composition over time (e.g., mergers and 
divestitures). When measuring performance, HCFA can account for these transitions by 
adjusting baseline levels or performance standards. 

c. Rewards and Penalties 

If the observed RPUPS in a performance year is less than the target, the rate of growth in 
average resource consumption is lower than the rate specified by the Federal government. 
Differences between expected and actual reimbursement rates for a provider's patients are 
deemed to be savings attributable to changes in relative efficiency. Multiplying by the number of 
Medicare patients involved in measuring the average reimbursements per patient produces an 
estimate of total Medicare Savings due to changes in the provider's relative efficiency. If 
RPUPS refers to professional and supplier services only (i.e., MVPS services), the estimated 
Medicare Savings will refer only to those services. RPUPS also could encompass all Medicare- 
covered services, and savings would be estimated accordingly. 



Figure ES-1: Deviations from Targets for 78 Providers in 10 Marl<et Areas: 1991 to 1992' 



J3 



3 
XI 

E 
cc 

I 

I 

E 
g 
m 

I -a 
I 

0) 



II 



-0.2 



0.4 



0.3 



0.2 



O.H- 



♦ t ♦ 

-0.1 ♦tt*>i'* 

-0.1*^ 



♦ ♦ 



-0.2 



♦ 



♦ > 



0.1 



0.2 



0.3 



0.4 



Percentage Deviations from Targets for Physician/Supplier Reimbursements 

Source: NCH file, 1991 and 1992. 



If the value of Medicare Savings is positive, the group has demonstrated improvement in 
relative efficiency. If the value is negative, the group has not met HCFA's budget goals. A 
value of zero means the group has met its target exactly; in other words, the group has met 
HCFA's budget goals for the average provider. These outcomes can be addressed in the payment 
system through policies that define rewards and penalties. 

In the proposed GVPS models, HCFA would retain its policy of setting uniform 
conversion factors for all phycicians. The GVPS pa}'ment system would distinguish between 
physician groups in terms of relative efficiency by giving lump sum reward payments to . 
successftil groups. At a minimum, the reward would be based on the actuarial value of any 
national fee penalties. HCFA could "refund" the value of lost revenues resulting from the 
national penalties to groups that operated within their targets. In addition, GVPS could result in 
rewards based on savings to Medicf.re and/or specific penalties for failing to meet performance 
standards. Penalty amounts could be withheld from future fee-for-service payments to the group. 

The value of additional rewards (and penalties) could be a function of several potential 
factors, beginning with the estimated Medicare Savings. We recommend that Medicare Savings 
be calculated on the basis of all Medicare services, not just physician and supplier services. It 
would be desirable and appropriate for Medicare to retain a portion of the savings. For reasons 
of equity and appropriate incentives, the physician group also could receive a portion of the 
savings, i.e., a reward. Because incentives and savings apply to reimbursements to all providers, 
one gauge for reward payments to the group is its PCR, i.e., the proportion of Medicare 
reimbursements for services provided to its patient population. The result would be "virtual 
capitation" to the group for its share of the patient population. Groups would have incentives to 
maximize savings and to increase their share of services provided to their patients. HCFA may 
choose other criteria for sharing savings (such as a specified Sharing Rate) instead of, or in 
addition to the Patient Capture Ratio. We recommend that rewards be equal to a refiind of the 
lost revenues due to reduced fee updates under MVPS, plus 75 percent of the product of the 
estimated savings and the group's PCR. 

Paying rewards for the success of groups raises questions about financing and budget 
neutrality. For savings and rewards related to MVPS services, HCFA still could use the national 
MVPS to calculate universal penalties. In those calculations, the Medicare Savings amount (for 
MVPS services) attributed to groups operating under GVPS could be added to the national 
aggregate expenditure totals. This would base national penalties on the performance of providers 
outside of GVPS. Providers under GVPS would receive the same conversion factors, reflecting 
the penalties, but the difference would be offset by the lump sum payments. 

For other types of services (i.e., not covered under MVPS), the Federal government also 
has processes for determining increases in payment rates. The context for making those 
determinations presumably includes budget considerations, although there is no structiire that is 



viii 



parallel to MVPS. HCFA may implicitly disregard estimated savings under GVPS when it 
makes these determinations, effectively creating a parallel situation in which the burden of 
financial penalties is concentrated on providers outside the umbrella of managed care under 
GVPS. 

d. Three Prototype Models 

Table ES-1 presents three model variations for GVPS. These three models are presented 
to illustrate important policy parameters regarding the scope of services and the incentive 
structure. Model 1 uses only professional and supplier services in the definitions of RPUPS and 
Medicare Savings, while Models 2 and 3 refer to all Medicare-covered services. Models 1 and 2 
carry reward potential only, with no specific penalties for failing to meet performance standards, 
whereas Model 3 carries potential rewards and penalties. We recommend Model 2 because we 
want to encourage participation by groups, and want to encourage the groups to manage total 
patient care. 

The formulas given in Table ES-1 for reward and penalty amounts are illustrative. In 
each of the models, the formulas refer to the entire scope of services used to define RPUPS. A 
model that includes provisions for specific penalties might share greater proportions of savings 
with the group. However, the formula for rewards and penalties need not be entirely symmetric. 
Furthermore, HCFA may decide to avoid or strictly limit the potential penalties, given the 
absence of patient "lock-in" provisions to control utilization. Model 1 would give rewards to 
groups equal to any refiands for reduced fee updates plus (Medicare Savings x Patient Capture 
Ratio X 0.75). As discussed above, multiplying by the PCR provides a usefiil estimate of the 
reimbursements that might have been paid to the group, had they not improved overall 
efficiency. The Sharing Rate of 0.75 represents a compromise between zero, which reflects 
traditional fee-for-service, and unity, which conveys a "virtual capitation" scenario. 

Model 2 would base reward payments on all Medicare-covered services, with amounts 
equal to any refiinds for reduced fee updates plus (Medicare Savings x Patient Capture Ratio x 
0.75). Model 3 carries higher potential rewards, equal to any refiands for reduced fee updates 
plus (Medicare Savings x Patient Capture Ratio x 0.95), and would account for negative 
Medicare Savings estimates through a penalty equal to (Medicare "Losses" x Patient Capture 
Ratio X 0.10). This would move toward a virtual capitation approach. 

5. Eligibility Criteria 

To a certain extent, criteria for eligibility depend on the purposes of GVPS. For basic 
purposes, such as enhancing equity in payment or applying incentives for efficiency as broadly 
as possible, HCFA may specify only a few criteria. These could include requiring formal 
contractual agreements among providers that operate under a single performance standard, 
specifying minimum sizes (e.g., number of beneficiaries served). We suggest criteria such as a 
minimum of 8,000 Medicare patients seen per year, a minimum PCR of 20 percent (for physician 



ix 



Table ES-1: Summary of Three Model Variations 




Model 1 


Model 2 


Model 3 


Services in 








Volume Measure: 


MVPS 


All Medicare 


All Medicare 


Formula for 








Sharing Savings: 








MVPS: 


(MS X PGR X 0.5) 


(MS X PGR X 0.75) 


(MS X PGR X 0.95) 


Other: 


None 


(MS X PGR X 0.75) 


(MS X PGR X 0.95) 


Formula for 








Penalties: 








MVPS: 


None 


None 


(MS X PGRx 0.10) 


Other: 


None 


None 


(MS X PGR X 0.10) 



MVPS services include most Part B professional and supplier services, and exclude 
outpatient department facility costs, ambulance services and durable medical equipment. 



MS refers to the estimated Savings for the group; PGR is the Patient Gapture Ratio, 
which is the proportion of all patients' Medicare reimbursements that were to the group. 



X 



services), and minimum Medicare reimbursements of $100,000 per year for evaluation and 
management physician services. 

HCFA may consider more restrictive criteria for two reasons. First, HCFA may want to 
limit the number of participants. This could lessen the overall administrative burden and 
concentrate resources on larger and more propitious sites. HCFA could undertake a definitive 
assessment of a group's capacity to manage the full range of Medicare-covered services. This 
may require several layers of criteria relating to the composition of the group, accessibility 
standards, quality of care standards, patient management practices, and information systems. 

6. Provider Responses 

If the Federal government made GVPS available, providers across the country would 
need to make a number of decisions in response. First, they could assess their interest and their 
ability to meet eligibility criteria. Most would have to pool their efforts with other providers in 
order to meet criteria relating to size and/or scope of services. 

Operating under GVPS, a group may make different decisions depending on 
characteristics of its patient population. For example, a group that focuses on primary care may 
have to ally itself with providers of more specialized care. A group that already provides a full 
range of services and a large portion of care to its patients may focus on managing the services it 
provides and expanding the number of Medicare patients it sees. Under the proposed models, 
groups would have incentives to control the volume and intensity of their own services, and to 
influence the services their own patients receive from other providers. 

7. Economic Consequences of GVPS 

We simulate the economic consequences of implementing GVPS on three parties: 
Medicare, GVPS groups, and other physician practices. Economic consequences result fi-om 
changes in total Medicare reimbursements for applicable services, plus any reward payments. 
The simulation model concerns the specific effects of GVPS on reimbursements for all Medicare 
services (Parts A and B) over a time frame of five years. 

Table ES-2 presents results of the simulation for the base case scenario, and contrasts 
them with projected results in the absence of GVPS. We assume that GVPS groups see 10 
percent of all beneficiaries that use services, realize lower utilization rates, and experience a 
slower than average rate of growth. Under GVPS, participating groups lose $1,988 billion in 
reimbursements in year 5 (12.92% of what they would have received without GVPS). This loss 
is more than offset by a reward of $2,509 billion for their success in meeting targets. With the 
reward, the groups are 3.38 percent better off with GVPS than without. The non-GVPS 
providers also face lower reimbursements under GVPS, but the losses are spread over a much 
larger base, and therefore only account for 2.73 percent of their year 5 reimbursements without 
GVPS. Finally, Medicare saves 2.44 percent of total reimbursements for year 5 with GVPS 
($7,709 billion), since higher payments to groups under GVPS are more than offset by lower 
payments to other providers. 



xi 



Table ES-2 

Distribution of payments with/without GVPS 
Scenario: GVPS Base Case* 



Payments ($M) in Year 5 


Scenario 


Difference (% change) with GVPS 


GVPS 


No GVPS 


Reimbursements to GVPS groups 
Reward payments to GVPS groups 


13,403 
2,509 


15,391 



-1,988 (-12.92) 
+2,509 


Total group income 


15,912 


15,391 


+521 (+3.38) 


Reimbursements to non-GVPS providers 


292,887 


301,116 


-8,230 (-2.73) 


Total payments by Medicare 


308,798 


316,507 


-7,709 (-2.44) 



* GVPS groups see 10% of beneficiaries that tise services in every year 



xii 



Table ES-3 examines the sensitivity of our results to various changes in the policy 
parameters. It may be seen that increasing the sharing rule from 75 percent to 95 percent reduces 
slightly the total payments by Medicare in year 5. However, it increases the groups' gain from 
implementation of GVPS, from 3.38 percent to 7.18 percent above their reimbursement total 
without GVPS. 

Of greater importance is the rebasing rule. The use of armual rebasing would make 
GVPS a money-loser for the groups, reducing their revenues 9.64 percent below the GVPS base 
case fifth year amount of $15,912 billion. This reflects the ratchet effect of continually adjusting 
targets based on actual performance. However, the groups' loss in this case is not a gain for 
Medicare. Instead, the benefits accrue to non-GVPS providers, who receive higher updates (and 
therefore smaller revenue losses) than they would otherwise. This is because rebasing reduces 
measured savings, and therefore reduces the rewards to GVPS groups which would otherwise be 
financed through lower updates. 

If the groups reduced their volume growth to 4.5 percent instead of 6.5 percent used in 
the base case, they would increase their income in year 5 by 1 .67 percent of the level the groups 
would receive if there were no GVPS. The income gain is smaller than the 3.38 percent achieved 
in the GVPS base case scenario. This suggests that the additional rewards for curbing utilization 
more tightly are ultimately outweighed by the loss of fee-for-service reimbursements. Using the 
4.5 percent growth assumption, Medicare would save 2.93 percent of year 5 payments without 
GVPS, compared to 2.44 percent savings in the GVPS base case with 6.5 percent utilization 
growth. 

Alternatively, if the GVPS groups increased their PGR by 2 percent per year in addition 
to achieving the baseline utilization savings for Medicare, they would greatly increase their fee- 
for-service reimbursements. In this scenario variation, the groups' revenues in year 5 would be 
29.3 percent higher than without GVPS. For Medicare, this scenario results in a 0.09 percentage 
point larger payment reduction than the GVPS base case because care is being transferred from 
non-GVPS to GVPS group providers, who are presumed to better control utilization grov^. 
Furthermore, if groups see 25 percent of beneficiaries. Medicare payments in year 5 are 
approximately $19.8 billion below their projected level with no GVPS program. This represents 
a 6.38 percent savings for Medicare. 

8. Conclusions and Recommendations 

Medicare has experienced large increases in spending during recent years, and most 
projections suggest these increases could continue. Policymakers have debated the merits of 
regulation, such as rate setting, versus market-oriented solutions, such as managed care. We 
believe that the regulatory framework established for MVPS is usefiil for achieving budgetary 
goals. However, we also believe it would be useful to supplement the current system with 
economic incentives that encourage the management of services. Many physicians and 
administrators who contributed to this study commented that incentives under the traditional fee- 



Table ES-3 



Effect of Varying Program Impacts on Payments Under GVPS 
GVPS Groups see 10% of Beneficiaries that Use Services 





Varying Assumptions 


Payments ($M) in Year 5 


Percent Change from No 
GVPS 


Scenario 


Groups' 
Utilization 
Growth 
Rate 


Annual 
Increase 

in 

Capture 


Groups' 
Share of 
Savings 


Annually 
Rebase 
Groups' 
Targets? 


To 
Groups 


To 
Others 


Total 


Groups 


Others 


Total 


No GVPS 










15,391 


301,116 


316,507 








GVPS Base Case 


6.5% 





75% 


No 


15,912 


292,887 


308,798 


+3.38 


-2.73 


-2.44 


Variants 




Vary Savings Share 


6.5% 





95% 


No 


16,497 


291,401 


307,898 


+7.18 


-3.23 


-2.72 


Vary Rebasing Rule 


6.5% 





75% 


Yes 


13,907 


296,665 


310,572 


-9.64 


-1.48 


-1.88 


Vary Utilization Growth 


4.5% 





75% 


No 


15,648 


291,570 


307,218 


+ 1.67 


-3.17 


-2.93 


Vary Capture Growth 


6.5% 


2% 


75% 


No 


19,900 


288,593 


308,494 


+29.30 


-4.16 


-2.53 



xiv 



for-service payment system — with or without MVPS — were out of step with their current efforts 
to manage care and improve efficiency. 

Managing services may result in lower utilization rates and large savings, but bringing 
beneficiaries into managed care envirorunents can be difficult. Medicare risk contracts with 
HMOs offer one opportunity. Capitation can create incentives for efficiency that encourage 
managed care. However, continuing participation by an HMO is largely contingent on positive 
financial results. In turn, enrollment by Medicare beneficiaries into an HMO is contingent on 
better benefits and/or lower premiums than competing Medicare supplemental policies. The 
current system pays HMOs 95 percent of estimated costs and therefore could save Medicar^^ up 
to 5 percent for enrollees. Unfortunately, Medicare saves less than 5 percent, or even loses 
money, in cases where the average cost estimates are too high because of favorable selection. 

We hypothesize that Medicare could achieve greater savings from GVPS than from the 
capitation system: 

► First, the chances of Medicare losing money are less under GVPS because the 
performance standards are based on the experience of the group. In contrast, capitation 
embodies "performance standards" that may have little correspondence to actual 
enrollees. Although there is always error associated with estimating expected costs, the 
experience of a group's own patients may be a more valid basis than the experience of 
other providers' patients. 

► Second, the financial benefits of managing care can be shared more evenly under GVPS. 
The formulas for sharing the savings can give ample incentives and rewards to groups, 
yet still allow Medicare to benefit substantially. Under capitation, any savings to 
Medicare are capped at 5 percent of mean reimbursement levels. Under GVPS, Medicare 
can keep the majority of savings for patients seen by most groups. 

► Third, under GVPS groups have incentives to serve and manage expensive Medicare 
patients. Providers paid under fee-for-service are encouraged to seek and retain patients 
most in need of services. Capitated health plans have incentives to seek and retain 
relatively healthy members, not patients. 

We also see advantages of GVPS over state-level MVPS and penalizing hospital medical 

staffs: 

► Our analysis suggests that increases in Medicare costs are more pronounced for Part A 
and other non-MVPS services. It would seem worthwhile to pursue comprehensive 
policies that embrace all services. Under GVPS, HCFA can follow reforms in the 
industry and encourage management of all services. Policies addressing MVPS services 
within states, or physician services within the hospital, are relatively narrow and not 
aimed at the major problem areas or most promising solutions. 



XV 

► Also worthwhile would be coherent polices that link appropriate incentives to the 
responsible decision-makers. We believe physician groups are the optimal focal points 
for comprehensive and coherent Medicare payment policies. Based on what we found, 
there are physician groups willing to accept the challenge. 

► In different ways, state-level MVPS and hospital medical staff policy options are subsets 
of potentially more comprehensive GVPS policies. Setting regional or local performance 
standards is one necessary step in establishing a GVPS option, which completes the 
process by giving ince^+ives to providers to respond. Hospital medical staffs are 
potential candidates for GVPS, which could give them incentives to manage ambulatory 
and institutional services. 

HCFA has several parameters to consider for GVPS, involving various tradeoffs. A 
significant decision is whether to stay with the scope of physician and supplier services. Other 
decisions have to do with the level of complexity to build into the algorithms for setting 
standards and measuring performance. Also, decisions are needed about the appiopriate balance 
between incentives to participate, and incentives for efficiency among those who participate. 

► HCFA has the administrative capacity and relevant data to implement GVPS for 
physician and supplier services, or for all Medicare services. Most of the administrative 
burden lies with the physician services because of their large numbers. Ironically, adding 
the other categories of services increases the administrative burden relatively little, but 
greatly increases the scope of the incentives and potential savings. We recommend 
basing GVPS on all services. For non-MVPS services, projections used in setting 
capitation rates for HMOs could also be used to set performance standards for groups. 

► There are a number of potential refinements and variations discussed in this report. 
Again, they would involve data and capacity that HCFA already has, but would add to the 
number of steps. The value of methods to dampen stochastic effects must be considered 
in light of selected criteria for participation. Allowing medium sized groups to 
participate, for example, may add to the value of refinements. We recommend that health 
status adjusters be employed, but further consideration is needed about which categories 
to use. 

► Groups should be given incentives for improving efficiency. These incentives could be in 
the form of rewards and/or penalties. Although penalties may strengthen incentives for 
efficiency, we believe that interest in participation would be greatly reduced by the 
prospect of losing money. Assuming that HCFA is willing to set cumulative performance 
standards, we recommend that concentrated penalties not be included. Failure to 
capitalize on an opportunity to manage care and earn rewards is itself a sufficient penalty, 
as is rising above a cumulative target and diminishing chances for future rewards. Giving 



xvi 

positive incentives similar to capitation, and allowing HCFA to share in the savings, 
could reap significant benefits for Medicare and participating groups. 

Based on these considerations, we recommend that HCFA consider models with 
parameters such as the following: 

► Establish eligibility criteria, such as groups with primary care physicians and specialists 
serving about 8,000 or more beneficiaries per year, and a PGR of at least 20 percent for 
physician services. 

► For GVPS beginning within the next year, say early in 1996, choose 1994 as the base 
year. 

► Measure utilization as RPUPS based on all Medicare services. 

► Establish performance standards using the counties where at least 5 percent of the group's 
patients reside. These are cumulative, meaning annual increases are applied to the 
previous target, not the most recent actual RPUPS. Adjust the performance standard each 
year for changes in case mix from the base year. 

► Pay successful groups a lump sum reward consisting of a refund for revenues lost from 
national fee penalties, plus 75 percent of the product of estimated Medicare Savings times 
the PGR. Pay rewards to a group for successfiil performance only if its cumulative 
Medicare savings, i.e., since beginning to operate under GVPS, are positive. 



Penalize unsuccessfiil groups only through any applicable national fee penalties. 



1 



I. Purpose 

The Health Care Financing Administration (HCFA) has periodically reformed Medicare's 
payment systems to help control costs. Price setting systems for hospital and physician services 
are notable examples. However, to control aggregate expenditures HCFA also must control the 
volume and intensity of services. Many payers have adopted managed care techniques including 
utilization management, selective provider contracting, and financial incentives. The policy 
options presented in this report could help propel the Medicare program into managed care and 
toward long run cost control. 

In 1989, the Federal govenmient initiated physician payment reform with legislation 
calling for the Medicare Fee Schedule (MFS) based on a Resource-Based Relative Value Scale 
(RBRVS). To address volume and intensity, the legislation established the Medicare Volume 
Performance Standard Rates of Increase (MVPS), which link physician fee updates to aggregate 
national Medicare expenditure outcomes for physician and supplier services. This research is 
intended to help HCFA explore alternatives to the current national MVPS, responding to three 
basic problems: 

► Physicians have weak economic incentives to be efficient because individual 
performance is aggregated with the rest of the nation. 

HCFA partitions services into separate volume performance standards for surgical, 
primary care and other services. However, Medicare-covered services delivered by all 
physicians in the country are aggregated into national pools. This approach gives weak 
financial incentives for providers to manage the volume and intensity of services because 
individual performance does not affect discemibly the national performance. 

► If national standards are exceeded, physicians who are relatively efficient get 
penalized as //"their own performance was average. 

In spite of the weak incentives under Medicare, some physicians practice more efficiently 
than others. This can result from differences in education, organization of practice, 
incentives given by other payers, etc. Medicare expenditure levels reflect the combined 
or average behavior of the relatively efficient and inefficient physicians. Blanket fee 
penalties do not differentiate between physicians in terms of their relative efficiency. 

► Over many years, shrinking Medicare payment levels relative to other payers may 
threaten beneficiaries' access to physician services. 

If the main mechanism for expenditure control is reducing physicians' Medicare fee 
levels, physicians may become less willing to accept Medicare patients. This unfortunate 
dynamic may occur without causing any general improvements in the cost-effectiveness 



2 



of service delivery patterns. Physicians' fees account for a distinct minority of total 
Medicare expenditures, but physicians make the decisions that affect most of the 
remaining services. 

This report discusses refinements to the national MVPS to allow qualified physician 
groups to elect separate annual performance standard rates of increase other than the national 
standard established for the year. Under the proposed models, called Group-Specific Volume 
' Performance Standards (GVPS), HCFA would track the annual reimbursement rates of Medicare 
beneficiaries seen by physician groups. The observed rates of increase would be compared to 
performance standards to determine whether or not the group was contributing to lower giuvvth 
rates in Medicare spending. 

This approach could encourage physicians to elect group-specific performance standard 
rates of increase and give incentives for improvements in efficiency. Moreover, incentives under 
GVPS can be expanded to all Medicare-covered services. Accordingly, physicians can be 
rewarded for managing all of their Medicare patients' services. 

n. Overview of the Project and Report 

A. The Project 

This is the final report for a three-year project intended to advise HCFA about separate 
performance standards for physician groups. The project has set out to: 

► Develop and analyze methods for measuring resource consumption at the provider level 
and for setting group-specific performance standards, 

► Develop and recommend criteria for determining whether a physician group is qualified 
to elect separate performance standards, 

► Consider the administrative requirements for HCFA to implement group-specific volume 
performance standards, 

► Describe how a physician group may operate under separate performance standards, and 

► Simulate policy implications of important model parameters. 

We have continued to develop conceptual models described in reports from previous 
projects sponsored by HCFA (Wallack et al., 1991 ; Tompkins et al., 1992). In the earlier work, 
we conducted empirical analyses of Medicare expenditure patterns for quasi-random samples of 
physician providers, using the Provider ID Number (i.e., billing number) to define providers. 



3 



For this project we combined many aspects of the model development and empirical 
analyses. We established an informal advisory committee consisting of representatives from a 
dozen physician organizations in different areas of the country. Through periodic meetings, 
conference calls and written correspondence these groups gave us usefiil reactions and insights 
throughout much of the project. In addition, we constructed data samples using these groups as 
main sites and other providers in the same market areas as additional observations in our 
secondary data analyses. 

The advisory committee included large multispecialty groups with reputations for 
delivering high quality of care and having strong management capability aimed at integration. 
The organizations that assisted us are: 

• Cleveland Clinic Foundation (Cleveland, OH) 

• Fallon Clinic (Worcester, Massachusetts) 

• Geisinger Clinic (Danville, PA) 

Henry Ford Health System (Detroit, MI) ^, 
Lahey Clinic Foundation (Burlington, Massachusetts) 
. • Lovelace Clinic (Albuquerque, NM) 

Mayo Foundation (Rochester, MN; Scottsdale, AZ; Jacksonville, FL) 

• Ochsner Clinic (New- Orleans, LA) 
Scott «fe White Clinic (Temple, TX) 

Upper Hudson Primary Care Consortium (Warrensburg, NY) 

Together, these organizations serve urban and rural market areas, and range in emphasis from 
primary care to national tertiary care referrals. Many also sponsor or contract with health 
maintenance organizations, including Medicare risk contractors. 

Our data samples were supplemented with other physician organizations chosen from the 
same market areas served by the advisory committee members. The Medicare National Claims 
History (NCH) file was our principal source of data. We retrieved claims records for all 
Medicare patients seen by providers in the sample during the years 1991, 1992 and 1993. In 
addition, "we obtained similar data for random beneficiary samples from the same market areas in 
order to estimate market level trends. 

B. The Report 

The first part of this report presents the purpose and aims of the study, and provides some 
background about the policy context for GVPS. Section III of this report briefly describes some 
other policy options to control expenditure levels. These include other potential refinements to 
the national MVPS, and HMO risk contracting. Section IV discusses the rationale and goals for 
a policy of Group-Specific Volume Performance Standards. These include objectives for HCFA 
and objectives that may exist for qualified providers. 



4 



The second part of the report discusses GVPS models for implementing performance 
standards at the group level. Also included are models for defining incentive structures based on 
the group's performance. Section V summarizes the basic approach to GVPS, and introduces 
key terms and relationships. The section describes ma-" "^odel variations together with related 
policy issues. Section VI describes methods for measuring resource consumption at the provider 
level. Empirical findings are presented for the sample providers. We consider the set of services 
currently under MVPS (i.e., professional and supplier services) and all Medicare-covered 
services. Section VII describes methods for setting performance standards. We investigate 
differences in growth rates between market areas, and growth rates at the provider level. Model 
options include year-to-year targets and cumulative targets projected from a specific base year. 
Section VIII considers specific changes to the payment system under GVPS that could affect the 
incentive structure for providers. As an alternative to the national MVPS, this approach could 
help to improve equity in payments and to encourage efficiency in service delivery. We consider 
the potential roles of economic rewards and penalties. 

The third part of the report discusses eligibility criteria for participafion under GVPS, 
potential provider responses to GVPS, and results from simulations of the economic 
consequences of GVPS. Section IX discusses criteria that HCFA may consider for assessing the 
qualifications of physician groups and their eligibility for participafion under separate 
performance standards. We begin with basic criteria relafing to the validity of ufilization 
measures and performance standards, and then discuss fiarther criteria related to potential policy 
goals of HCFA for GVPS. Section X anficipates how providers may respond if GVPS were 
implemented. Groups must consider whether to elect a separate performance standard and, if so, 
how to manage for success. Section XI presents the simulafion results, comparing economic 
outcomes for Medicare, for successful groups within GVPS, and for physicians who are not 
operating under separate performance standards. Results from this first generation simulation 
model are intended to help facilitate discussion about fiarther research and the implications of 
particular policy choices. Section XII contains conclusions and recommendations for GVPS 
models. 

The report also contains several appendices. Appendix A provides further information 
about the physician groups comprising our advisory committee. Appendix B describes the 
construction and handling of data files used for the empirical research tasks. Appendix C 
describes potential refinements of the utilization measures and performance standards presented 
in this report. 

III. Policy Context and Alternatives 

This project describes the policy of GVPS, perhaps one of several policies HCFA may 
consider in addressing the following general problems faced by the Medicare program: 



5 



► Increasing costs. A fundamental problem is the rising cost of financing Medicare- 
covered services. 

► Incentives to provide more services. Expenditixre increases are driven in part by 
prevailing fee-for-service incentives for providers to deliver more services than may be 
necessary. 

" ► Potential threats to access and quality. Policies designed to control costs can lead to 
concerns about reductions in quality of care, access to needed services, and the liberty to 
choose among providers. 

This section briefly reviews three other policy alternatives against which HCFA may 
weigh the relative merits of GVPS: state level MVPS, HMO risk contracting, and focusing on the 
volume and intensity of services delivered by hospital medical staffs. 

A. State Level MVPS 

An alternative to the national MVPS would be to mandate geographic groupings of 
physicians, such as a region or state (Holahan and Zuckerman, 1993; PPRC, 1990, 1992). There 
are at least two potential benefits of moving in this direction. First, there are observable 
differences between market areas in expenditure growth rates. Measuring the volume 
performance of different states separately could give greater relevance to the target for providers 
than average national rates. Second, states may develop infrastructures as part of health reform 
efforts that create the ability to monitor and/or enforce changes in practice patterns at more local 
levels. 

There are caveats and limitations associated with this policy proposal. Absent the 
developmeni of infrastructures under more elaborate health care reforms, states or regions are not 
likely to have mechanisms to monitor or cause changes in physicians' practice patterns.' Also 
relevant to this approach is the issue of patients crossing state borders to receive services. 
Finally, applying reduced fee updates for failure to meet performance standards could create 
price differentials across state borders: physicians in successful states would trend toward higher 
Medicare payment rates and beneficiary copayment rates than physicians in unsuccessful states. 



' Physicians also may be differentiated by specialty. Specialty societies currently exert some 
"soft" influence on their members, especially regarding issues of clinical appropriateness. 
Extending this influence to counteract economic incentives may or may not be feasible. 



6 



B. HMO Risk Contracting 

For over a decade, HMOs have had the opportunity to contract with HCFA to provide all 
Medicare-covered services for beneficiaries who choose to enroll. Generally, capitation 
promises several advantages over fee-for-service. For the provider, the combination of 
enrollment and fixed capitation payments permits control over the development and operation of 
an integrated service delivery system. The payer knows its costs in advance, essentially 
delegates the responsibility for managing care to the contracting health plan, and generally 
oversees the process. Also appealing about capitation is the overlapping interest in reducing 
utilization: The health plan can earn financial profits by lower utilization rates, while the payer 
may reduce its projected expenditure growth rates. 

Recently, many HMOs have shown interest in Medicare risk contracting, and the trend is 
toward higher aggregate enrollment. This is consistent with a deepening penetration of managed 
care organizations in many health care markets. Many providers and insurance companies pre 
entering multiple contracts to help assure continued or expanded market shares. As in the early 
1980's, there is a widespread sentiment that capitation is the "wave of the future" for most 
payers, including Medicare. 

There are difficulties surrounding HMO risk contracting. First, enrollment patterns with 
favorable selection virtually preclude savings for Medicare. This is because the payment system 
has weak adjustments for differences in health status between HMO enroUees and beneficiaries 
in the fee-for-service sector. Although HMOs also can experience unfavorable selection, the 
voluntary nature of the system allows those plans to exit the Medicare market, leaving only 
HMOs that make money (Porell and Tompkins, 1993). Second, under capitation the health plan 
reaps profits from lower utilization while Medicare keeps relatively little of any savings 
generated through efficiency. Managed care can lead to lower costs (e.g., fewer hospital days of 
care); however, research findings do not indicate strongly that capitation is necessary to achieve 
lower costs (Miller and Luft, 1994). 

C Penalize High Cost Hospital Medical Staffs 

This proposed approach would withhold a portion of payments to physicians who 
practice in hospitals deemed to deliver the highest average volume and intensity of service per 
admission (Welch and Miller, 1994). HCFA would return the amount withheld if the physicians 
could reduce their average relative value units per admission. This approach is attractive in part 
because it gives responsibility to physicians who have the authority to manage the relevant 
utilization patterns. Financial penalties would be focused more on providers identified to have 
high costs, rather than applying penalties uniformly to all physicians. 

A difficulty with this policy is that it focuses on specific episodes and not necessarily on 
managing all patient care. In addition, it might offer perverse incentives to shift the provision of 
care away from the inpatient setting, even if unwarranted. Another possible drawback is the 



7 



regulatory and adversarial approach to physicians. While engendering goodwill in the physician 
community need not be HCFA's primary objective, many other payers are concluding that 
working with physicians is their preferred strategy for meeting long rvin goals of efficient service 
delivery. 

IV. Rationale and Goals for GVPS 
A. Rationale 

The Federal government has implemented MVPS to control growth in expenditures for 
professional and supplier services. Payments exceeding target levels can be recouped by 
reducing updates to fees (i.e., the conversion factors) under the Medicare Fee Schedule.' This 
provides HCFA with a useful tool for enforcing predetermined budgets. However, pooling all 
physicians' reimbursements at the iiational level gives weak "collective" incentives for 
physicians to control utilization levels. Attempts by individual physicians to become more 
efficient are unrewarded. Even worse, efficient providers who do not provide unnecessary 
services are penalized immediately through lower revenues. 

The Medicare physician payment system involves a dual approach to cost control: strong 
regulation of prices, and weak incentives to control volume and intensity. There were concerns 
that strong price regulations could lead physicians to increase volume and intensity in order to 
offset reductions in their personal incomes. Physicians are aware of these incentives to inflate 
volume ("churning patients") in order to realize target income levels. In the future, policymakers 
may want to replace this simple "treadmill" with policies that evoke appropriate responses from 
physician organizations. Specifically, we are aiming to encourage strategic orientations that 
focus on managing all services received by patients. 

Medicare is one among many payers in the U.S. health care system. Ove; time. Medicare 
physician fee levels could become low enough to hinder beneficiaries' access to services. Some 
analysts are concerned that Medicare may be approaching that point already (PPRC, 1994). To 
steer away from this possibility. Medicare needs coherent payment policies that allow physicians 
to manage total expenditures through cost-effective delivery systems, and do not rely on just low 
prices per unit of service. 

Congress acknowledged that refinements to the basic national approach to expenditure 
control could be warranted, and specifically called for development of group-specific 
performance standards. The legislation that initiated MVPS included the following provision: 



' We use the terms volume performance standards and targets interchangeably. 



8 



... the Secretary shall . . . implement a plan under which qualified 
physician groups could elect annually separate performance 
standard rates of increase other than the [national] performance 
standard rate of increase established for year . . . The Secretary ^ 
shall develop criteria to determine which physician groups are 
eligible to elect to have applied to such groups separate 
performance standard rates of increase and the methods by which 
such group-specific performance standard rates of increase would 
be accomphshed. (OBRA 1989, Section 1848f 4) 

This project is intended to help HCFA explore this possibility. 

B. Goals for HCFA and Providers 

HCFA conducts business with providers whenever beneficiaries receive Medicare- 
covered services. HCFA views transactions as expenditures, while providers view them as 
revenues. In the aggregate, successful control of Medicare expenditures translates into reduced 
revenues for providers. This creates an inherent divergence of interest between HCFA and 
providers generally. Still, the potential overlap of interest is very large as well, such as the 
well-being of beneficiaries and the quality of care. The size of this overlap relative to the 
divergence of interest is greatest between HCFA and providers that provide the highest quality of 
services, while being relatively efficient. 

We assume HCFA would espouse the following four potential goals for a comprehensive 
physician payment policy: 

► Achieve and maintain high quality health care. Medicare beneficiaries should receive 
services that equal or exceed prevailing quality standards in the local community. 

► Improve clinical efficiency. For each beneficiary, the combinafion of services results in 
the lowest cost necessary to meet the quality standards. This means avoidance of 
unnecessary referrals, tests, and services. 

► Improve market efficiency. This is multifaceted: 

• Relatively efficient providers are rewarded and increase their market share, 

• Inefficient providers bear the brunt of fee reductions or other penalties, 

• Providers generally become more efficient in response to the market dynamics, or 
lose market share. 



9 



► Achieve budget neutrality. Implementing GVPS does not lead to higher aggregate 
Medicare expenditure levels. 

Providers electing to participate under GVPS may not all have the same objectives or 
priorities. However, participation might help them achieve particular goals: 

► Compensation for clinical efficiency. Savings attributable to clinical efficiency could 
be shared with participating providers. 

► Greater market share. This may be achieved by increasing the number of patients 
served and/or the scope of services provided. 

► Support for new strategic orientations. Market pressures for efficiency create new 
challenges and new opportunities for providers. The data and potential for rewards 
associated with GVPS could promote strategic orientations that include, but go beyond 
Medicare. 

GVPS models may not be better in every respect than the alternatives described in 
Section III. However, they do resemble the alternatives in certain important ways: 

► Managed care. GVPS models could put physician organizations "in the driver's seaf ' 
with respect to managing a broad range of services delivered to their patients. As with 
HMOs, this policy moves in the direction of integrated delivery systems that manage total 
patient care. 

»■ Involvement of high cost beneficiaries. Aggregate Medicare expenditure levels are 
driven by services delivered to a small minority of beneficiaries. Like the high cost 
medical staff policy proposals, GVPS could engage the providers who currently serve 
high cost beneficiaries. This is a distinct advantage over capitation, which encourages 
participation by health plans that do not enroll high cost patients. 

► Focused individual and collective incentives. GVPS could change incentives for 
providers who elect separate performance standards. GVPS also could concentrate 
penalties and collective incentives on the remaining physician subpopulation — similar to 
the goal of state level MVPS options. 

V. Summary of Basic GVPS Models 
A. The Policy Context: MVPS 

We propose policy options based on Group-Specific Volume Performance Standards 
(GVPS). This approach would supplement and refine the national MVPS, which focuses on 



10 



rates of increase rather than absolute expenditure levels. This is consistent with policy goals of 
achieving sustainable grov^h rates in Medicare spending. Proposed GVPS models emulate the 
focus on growth rates, which could encourage participation by providers with relatively sick 
patients and/or elaborate practice styles. Over time, HCFA could transition to models that 
evaluate providers on their expenditure levels relative to other providers, as well as grov/th rates. 

GVPS models must depart from some aspects of the current MVPS approach. First, 
national expenditures are measured in total, with adjustments for changes in the Medicare 
population size. At the provider level, it is more valid and convenient to measure performance 
on average for patients seen. Second, current performance standards are set nationally and 
reflect average performance not only across providers but also across market areas. Valid 
performance standards for providers need to reflect local market conditions, not expectations 
averaged across all market areas in the country. Third, the case mix of patients seen by a 
provider can change over time and performance standards should account for such changes. 
HCFA has experience with these issues through its capitation payment system, and we borrow 
from some methods already used by HCFA. Finally, GVPS could encourage physicians to 
manage all Medicare services for their patients. Thus, the policy could embody a vision for 
physicians to manage their patients, not just their own practices or only professional and supplier 
services. 

B. Methods to Monitor Volume Performance 

We measure average resource consumption at the provider level by Reimbursements Per 
Unique Patient Seen (RPUPS). In the denominator of this ratio are the Medicare beneficiaries 
who receive physician services from the group during the calendar year. In the numerator are 
Medicare reimbursements to all providers seen by these beneficiaries during the calendar year, 
for the services included in the measure. Two alternative scopes of service are contemplated for 
the GVPS option: 

• Those services currently under MVPS. 
• . All Medicare-covered services. 

In either case, providers are reimbursed for services according to all of the applicable prevailing 
payment policies, including the Medicare Fee Schedule for physician services, the hospital 
Prospective Payment System, etc. 

RPUPS G Y = (Reimbursements to All Providers for Patients of G) y ^ (N) g, y 

where G is a particular group operating under its ovra GVPS, and 

Y is a calendar year, either a base year or a performance year. 



11 



Under GVPS, HCFA would compare resource consumption in a given year to a target 
level or performance standard that is derived from resource consumption levels for a base year, 
times a specified percentage rate of increase. 

Target = Expected RPUPS performance Year 

= RPUPS Base Year ^ Rstc of Incrcasc 

where the Rate of Increase is specified by the Federal government. 

Targets could be set on a year-to-year basis, using the most recent observed RPUPS as a base. 
Preferably, targets could be updated cumulatively from the level of RPUPS observed in a 
specified base year, without regard to intermediate values of RPUPS. 

If the observed RPUPS Penonnunce Year is less than the target, the rate of growth in average 
resource consumption is lower than the rate specified by the Federal government. We explore 
methods to adjust for differences in a provider's case mix between the base year and performance 
year. Differences between expected and actual reimbursement rates for a provider's patients are 
deemed to be savings attributable to changes in relative efficiency. Multiplying the savings 
amount per patient by the number of Medicare patients seen by the provider produces an estimate 
ot total Medicare savings due to changes in the provider's relative efficiency. 

Medicare Savings y = (Target' y - RPUPS y ) x N y 

where Y is a given year in which a group's performance is being evaluated, 

Target' is the predicted RPUPS y, adjusted for differences in the provider's case 
mix between the base year and the performance year, and 

N is the number of Medicare beneficiaries who received physician services firom 
the group (i.e., the denominator in RPUPS y)- 

If RPUPS refers to professional and supplier services only (i.e., MVPS services), the estimated 
Medicare Savings will refer only to those services. RPUPS also could encompass all Medicare- 
covered services, and savings would be estimated accordingly. 

In addition to measuring Medicare savings each year, HCFA may choose to evaluate 
Cumulative Medicare Savings attributable to a group's performance. In that context, each year 
the group will make a positive or negative increment to its cumulative savings. Whether a group 
has accumulated a surplus or a deficit may affect the economic consequences of Medicare 
Savings in any given year, in terms of rewards or penalties. 



12 



C. Rewards and Penalties 

If the value of Medicare Savings in a given year is positive, the group has demonstrated 
improvement in relative efficiency. If the value is negative, the group has not improved its 
relative efficiency that year. A value of zero means tne group exactly has met its target; in other 
words, the group has met HCFA's specified growth rate for the average provider. These 
outcomes can be addressed in the payment system through policies that define rewards and 
penalties. Here we briefly present the concepts and potential mechanisms for dealing with 
rewards and penalties. 

Under MVPS, failure to meet national performance standards leads to blanket penalties 
for all physicians. The mechanism is to reduce fliture increases to physicians' fees. In the 
proposed GVPS models, HCFA would retain its policy of setting uniform conversion factors for 
all physicians. The GVPS payment system would distinguish between physician groups in terms 
of relative efficiency by giving lump sum reward payments to successful groups. The most 
conservative type of reward would be based on the actuarial value of any national fee penalties: 
HCFA could "refund" the value of lost revenues. We also consider models that would go 
beyond recompensing groups for fee penalties. GVPS models could give rewards based on 
savings to Medicare, and/or specific penahies for failure to meet performance standards. Penalty 
amounts could be withheld from future fee-for-service payments to the group. 

The value of additional rewards (and penalties) could be a function of several potential 
factors, beginning with the estimated Medicare Savings. It would be desirable and appropriate 
for Medicare to retain a portion of the savings. For reasons of equity and appropriate incentives, 
the physician group also could receive a portion of the savings, i.e., a reward. Because incentives 
and savings apply to reimbursements to all providers, one gauge for reward payments to the 
group is its proportion of Medicare reimbursements for its patient population. We call this the 
Patient Capture Ratio (PCR). HCFA may choose other criteria for sharing savings, instead of or 
in addition to the capture ratio such as a simple rule or Sharing Rate. For example, HCFA may 
choose simply to share savings equally with the group on the premise that all appropriate 
Medicare savings are equally valuable to HCFA, whether they resuU in lower fee-for-service 
payments to the group or to other providers.^ Using these criteria, general models for rewards 
and penalties are: 

Reward Group = Refund + (Medicare Savings x Patient Capture Ratio Group ^ Sharing Rate) 
Penalty Group = (Medicare Losses x Patient Capture Ratio Group ^ Sharing Rate) 



^ The values of the Sharing Rate may differ for rewards and penalties. Also, HCFA could 
specify limits on the values of any reward or penalty. 



13 



where rewards (penalties) may occur only if a group has positive (negative) Cumulative 
Medicare Savings. 

Paying rewards for the success of groups raises questions about financing and budget 
neutrality. For savings and rewards related to MVPS services, HCFA still could use the national 
MVPS to calculate universal penalties. In those calculations, the Medicare Savings amount (for 
MVPS services) attributed to groups operating under GVPS could be added to the national 
aggregate expenditure totals. This would ignore the savings generated by the GVPS providers 
and base penalties on the performance of providers outside of GVPS. Providers under GVPS 
would receive the same conversion factors, reflecting the penalties, but the difference would be 
offset by the lump sum payments.^ 

For other types of services (i.e., not covered under MVPS), the Federal government also 
has processes for determining increases in payment rates. The context for making those 
determinations presumably includes budget considerations, although there is no structure that is 
parallel to MVPS. HCFA may implicitly disregard estimated savings under GVPS when it 
makes these determinations, effectively creating a parallel situation in which the burden of 
financial penalties is concentrated on providers outside the umbrella of managed care under 
GVPS. 

Table 1 presents three model variations for GVPS. These three models are presented 
from among the many potential variations to illustrate important policy parameters regarding the 
scope of services and the incentive structure. Model 1 uses only professional and supplier 
services in the definitions of RPUPS and Medicare Savings, while Models 2 and 3 refer to all 
Medicare-covered services. Models 1 and 2 carry reward potential only, with no specific 
penalties for failing to meet performance standards, whereas Model 3 carries potential rewards 
and penalties. 

The formulas given in Table 1 for reward and penalty amounts also are illustrative. In 
each of the models, the formulas refer to the entire scope of services used to define RPUPS. A 
model that includes provisions for specific penalties might share greater proportions of savings 
with the group. However, the formula for rewards and penalties need not be entirely symmetric. 
Furthermore, HCFA may decide to avoid or strictly limit the potenfial amounts of any penalfies, 
given the absence of patient "lock-in" provisions to control utilization. Model 1 would give 
rewards to groups equal to any reftinds plus (Medicare Savings x Patient Capture Ratio x 0.75). 



^ In an alternative discussed later, HCFA would add back only the specific reward payments 
to groups in the calculations for MVPS. This would partially credit the nafional performance 
with the success of the groups. 



14 



Table 1: Summary of Three Model Variations 




Model 1 


Model 2 


Model 3 


Services in Volume 
Measure: 


MVPS 


All Medicare 


All Medicare 


Formula for Sharing Savings: 


MVPS: 


(MS X PGR X 0.75) 


(MS X PGR X 0.75) 


(MS X PGR X 0.95) 


Other: 


None 


(MS X PGR X 0.75) 


(MS X PGR X 0.95) 


Formula for Penalties: 


MVPS: 


None 


None 


(MS X PGR X 0.10) 


Other: 


None 


None 


(MS X PGR X 0.10) 



MVPS services include most Part B professional and supplier services, and exclude ASC and 
outpatient department facility costs, ambulance services and durable medical equipment. 



MS refers to the estimated Medicare Savings for the group. PGR is the Patient Gapture Ratio, 
which is the proportion of all patients' Medicare reimbursements that were to the group. 

As discussed above, multiplying by the PGR provides a useful estimate of the reimbursements 
that might have been paid to the group, had they not improved overall efficiency. The Sharing 
Rate of 0.75 represents a compromise between zero, which reflects traditional fee-for-service, 
and unity, which conveys a "virtual capitation" scenario. 

Model 2 would base reward payments on all Medicare-covered services, with amounts 
equal to any refunds plus (Medicare Savings x Patient Gapture Ratio x 0.75). Model 3 carries 
higher potential rewards, equal to any refunds plus (Medicare Savings x Patient Gapture Ratio x 
0.95), and would account for negative Medicare Savings estimates through a penalty equal to 
(Medicare "Losses" x Patient Gapture Ratio x 0.10). This moves toward virtual capitation. 



15 



VI. Utilization Measures 

A. The Basic Measure: RPUPS 

HCFA monitors total reimbursements for physician and supplier services under MVPS, 
making expHcit allowance in the performance standards for growth in the Medicare population. 
Under GVPS models, however, we work with average reimbursements because the number of 
patients seen by a group can change unpredictably over time. Similarly, insurers often express 
their costs as rates per person, "per member-month," etc. These are convenient measures for 
comparing differences between subpopulations, and for assessing trends over time even when the 
population size changes. In the GVPS context, the "population" is the set of Medicare 
beneficiaries seen by a physician organization during a calendar year. Costs are measured from 
HCFA's perspective, i.e., as Medicare reimbursement amounts for patients seen by the group. 

One goal for GVPS is to help control aggregate reimbursements, or equivalently, 
reimbursements per beneficiary. A physician group that manages utilization within its own 
system is likely to contribute to lower Medicare costs overall. However, an objective here is to 
develop an appropriate measure that takes account of any excessive or potentially offsetting 
services from other providers. Therefore, the measure of provider performance includes all 
relevant services for patients seen, including those delivered by other providers. We would like 
physician groups to embrace the perspective of the entire patient, for clinical and economic 
reasons. 

The ratio constructed from the unique Medicare patients seen by a provider and those 
patients' Medicare reimbursements, we call Reimbursements Per Unique Patient Seen (RPUPS). 
This measure includes only services delivered to a provider's patients — not all beneficiaries in 
the same city, county, state, etc. Accordingly, the measure of an individual provider's 
performance would exclude beneficiaries who either saw no physician during the period, or saw 
only physicians outside that physician organization. 

There are several potential variations on the basic definition of RPUPS. Whatever the 
specific definition, RPUPS would be defined and measured the same way each year. RPUPS for 
one year would constitute the baseline to which a target rate of increase is applied. Comparing 
changes in the value of RPUPS between years to target rates of change will allow HCFA to 
determine whether a physician organization is helping to achieve national aggregate target rates 
of increase. 

B. Scope of Services 

MVPS includes most Medicare Part B physician, professional and supplier services. 
Among services not included are facility usage (including ambulatory surgical centers and 
hospital outpatient departments), ambulance services, and durable medical equipment. The same 
set of services could be applied to the utilization measure for physician groups under GVPS. 



16 



Alternatively, the scope of services could be expanded to include institutions and all other 
services. We consider models that differ in terms of the scope of services included in the 
measure of physicians' performance: 

• Services currently included under MVPS, or 

• All Medicare-covered services. 

Expanding the roster of services could help GVPS to extend incentives for efficiency to 
the rest of Medicare-covered services. Inclusion of other services could give more opportunities 
for physicians to reduce overall expenditures. It is believed that HMOs have achieved savings 
relative to fee-for-service through reduced hospital utilization. This may continue to be true, 
although as hospital days of care decline in most regions, management of home health care, 
post-acute, and professional services becomes more important in competitive markets. In 
general, the successes of managed care generally are likely to stem from cost-effective 
innovations and substitutions involving a wide range of services. 

During this study we worked with physician groups located in different parts of the 
country. Our study files contain data from the geographic areas where most of these physician 
groups are located. We defined the geographic areas as the three-digit ZIP Code areas that 
account for at least five percent of Medicare patients seen by the original set of groups. From 
each of the study areas, we selected a sample of providers that appeared to be the most likely 
candidates for GVPS. Selection was based on the provider having a large number of physicians, 
a large number of beneficiaries seen, and high Medicare reimbursement levels in 1992. 

Also, from each area we drew random samples of beneficiaries who received services 
from any physician provider in the selected geographic areas, in order to determine market 
trends. We obtained all Medicare claims records from 1991 through 1993 for every beneficiary 
who saw any of the sample providers anytime during that period. The average number of 
Medicare beneficiaries seen at least once in 1992 by the 78 providers was about 1 5,000. For the 
beneficiaries sampled randomly we also obtained all Medicare claims records from the years 
1991 through 1993. See Appendix B for a fuller description of the data steps. 

C. Illustrative Findings 

Figures 1 and 2 show the values of RPUPS in 1992 for 78 selected providers located in 
ten of the geographic areas included in this study. The values of RPUPS differ substantially 
between providers, reflecting differences in case mix, relative efficiency, input prices, and 
Medicare carrier or fiscal intermediary practices. 

Figure 1 shows values of RPUPS based on reimbursements for physician and supplier 
services. For all 78 providers, RPUPS averaged $1,969 and ranged from $706 to $5,200. If the 
providers with the single highest and single lowest extreme values are ignored, RPUPS averages 



Figure 1: RPUPS Values for 78 Providers Selected from 10 Markets, Physician and Supplier 

Services 



6000 



5000 



4000 



J 3000 

o 

Q 



2p00 



1000 



I 



l"i' MMM 'i 



ilJ| U |lljlli »| LI|l.lil.l l.l| U| l.l|l.l|» | ll ll,l.l|l.l|l.l|N|l.l|ll|l.l,IJ,U, 



i " i- r 'i 



i "r' i "i" i 



1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 



Source: National Claims History file, 1992. 

Note: Reimbursements Per Unique Patient Seen (RPUPS) are the mean Medicare reimbursements for Medicare patients seen 
by the provider, unadjusted for case mix or geographic differences in prices. 



Provider 



Figure 2: RPUPS for 78 Providers Selected from 10 Markets, All Medicare Services 



00 



25000 



20000 



15000 



a 

"3 
O 



10000 



5000 



|UiU|li|li|ll|llilJjUi 



1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 ^5 57 59 61 63 65 67 69 71 73 75 77 

Provider 

Source: National Claims History file, 1992. 

Note: Reimbursements per Unique Patient Seen (RPUPS) are the mean Medicare reimbursements for Medicare patients seen by the provider, 
unadjusted for case mix or geographic differences in prices. 



19 



$1,944 and ranges from $759 to $3,907. Figure 2 shows results for all Medicare-covered 
services. For all 78 providers, RPUPS averaged $8,297 and ranged from $3,072 to $24,451. If 
the providers with the single highest and single lowest extreme values are ignored, RPUPS 
averages $8,153 and ranges from $3,324 to $16,647. 

Figure 3 shows the proportion of total reimbursements that was for physician and supplier 
services for each provider's patients. Physician and supplier services averaged 25 percent of the 
total and ranged from 15 to 39 percent. Figure 4 shows that there was also a wide variation in 
the proportion of reimbursements for physician and supplier services that were paid to that 
provider, averaging 1 8 percent and ranging from 2.9 percent to 52.9 percent. This represents the 
Patient Capture Ratio for those services. 

Physician groups were defined using the Provider Tax Number, also known by the IRS as 
the Employer Identification Number (EIN), which is available only for physician and supplier 
service claims. However, physician group practices can be part of integrated health systems that 
include facilities and other types of providers. Data limitations (i.e., the absence of EINs) make 
it difficult to measure capture ratios for these other types of services. Calculating the PCR for all 
Medicare services would require the Provider Identification Numbers of facilities and other 
provider entities affiliated with the physician group. Later we discuss some possible 
implications of these alternatives (see Section VIII, E). 

For GVPS, relevant comparisons relate to changes in RPUPS over time in relation to 
changes in market- wide measures of RPUPS. We discuss those comparisons next, along with 
setting targets and estimating savings due to relative efficiency. Note that Appendix C presents 
potential variations in the definition of RPUPS, including removing some beneficiaries from 
RPUPS, removing some categories of reimbursement from RPUPS, and adjusting RPUPS for 
changes in case mix. 

VII. Setting Performance Standards or Targets 

A". Scope of Services 

HCFA can calculate the value of RPUPS each year for every group operating under 
GVPS. In order to estimate the change in relative efficiency of the provider, as well as the 
resulting savings to Medicare, we need to set a target, or volume performance standard, for the 
provider. As discussed above, we explored measures of RPUPS based on physician and supplier 
services, and all Medicare-covered services. 

The most straightforward implementation of GVPS, from a regulatory point of view, 
might involve just the services under the current MVPS (i.e., most physician, professional and 
supplier services). The Federal government already specifies rates of increase for those services. 



Figure 3: RPUPS Values for Physician and Supplier Services, as a Percentage of RPUPS Values 

for All Services 




1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 56 57 59 61 63 65 67 69 71 73 75 77 

Provider 

Source: National Claims History file, 1992. 

Note: Reimbursements per Unique Patient Seen (RPUPS) are the mean Medicare reimbursements for Medicare patients seen by the provider, 
unadjusted for case mix or geographic differences in prices. 



Figure 4: Reimbursements to Provider vs. All Physician and Supplier Reimbursements 

(Patient Capture Ratio) 



0.6 



0.5 



0.4 



0.3 



0.2 



0.1 



|l, l | U |l. l | U|U jU|U,U|l.l|U i LI|l.l|l,l|l, l|l l | l,l| l .l|IJ|Ui 



|I.I| U |I,I| M| I,I | I, I| I,I|I,I| I ,I|1, I | I .I,I,I| 



1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 .57 59 61 63 65 67 69 71 73 75 77 



Source: National Claims History file, 1992. 



Provider 



22 



and GVPS could be adapted from The national performance standards. One difference would be 
to remove that portion of the total increase attributable to growth in the number of Medicare 
beneficiaries, since volume is measured per person under GVPS. 

However, the GVPS alternatives would allow HCFA to introduce incentives for providers 
to manage care. Consideration needs to be given to the appropriate scope of services. Health 
care delivery systems of today often are more vertically integrated and include primary care, 
tertiary care, facilities, etc. For hospital and other institutional services, HCFA would need to 
expand the MVPS concept. Presumably, HCFA would set targets that were commensurate with 
the standards for physician and supplier services. In other words, similar actuarial principles or 
policy objectives would be embodied in the methods and data used for the different categories of 
services. HCFA could use projections that are involved in setting Adjusted Average Per Capita 
Cost (AAPCC) rates for Medicare payments to HMOs. 

Although MVPS is now limited to physician and supplier services, more substantial cost 
increases are occurring for other services (see discussion below). This adds another practical 
advantage to expanding GVPS to include all Medicare-covered services, and to give physicians 
incentives to manage the whole range of services for their patients. 

B. Differential Growth Rates Based on Location 

Under MVPS, the Federal government specifies national rates of increase for the coming 
year. The current MVPS does not explicitly account for geographic variations in the factors 
underlying the national performance standards. Targets are set and performance is monitored at 
the national level only. This is consistent with MVPS being a budget tool and not an incentive 
system. Total spending for those services is governed, even though actual rates of increase can 
differ among regions and local areas. Between 1990 and 1991, expenditures per Medicare 
enroUee for physician and supplier services actually fell in several states, and grew by as much as 
21 percent (in Nebraska). The nation averaged a 4-percent increase during this period (PPRC, 
1993). 

Moving to genuine incentive systems may require that more consideration be given to 
appropriate geographic differences in the target rates of increase. A national target is a weighted 
average of local differences, and does not conform to the local perspectives of states, coimties, 
cities, or providers. The target rate of increase should reflect a level of expectation that is 
consistent across sites in different geographic areas. A provider should not win or lose under 
GVPS simply because of its location. 

At the other extreme, target rates of increase perhaps should not be based solely on the 
historical experience of the provider. Basing rates on lower (higher) rates of growth by a 
provider would tend to penalize (reward) historical efficiency (inefficiency). A challenge vmder 
GVPS will be to define actuarial reference populations that reflect relevant local trends but do 
not focus too closely on the provider's own experience. In this study we operationalize the 



23 



concept of targets using data in our study files. We compare average increases within markets 
areas to rates of change at the provider level. This allows us to examine the stability and rates of 
change in RPUPS at the provider levels. 

Table 2 shows the rates of growth in Medicare reimbursement rates per patient for the 
random samples of beneficiaries drawn from each of the ten market areas." Generally, there were 
much greater increases for all services compared to physician and supplier services, and greater 
increases between 1991 and 1992 than between 1992 and 1993. For physician and supplier 
services from 1991 to 1993, larger increases occurred in Massachusetts (8%), Michigan and 
Peimsylvania (4%), Florida (3%), and Minnesota (2%). Lower increases occurred in Ohio (-2%), 
Arizona (0%), Louisiana (1%) and New Mexico (1%). Between 1991 and 1993, larger increases 

Table 2: Rates of Change for Medicare Reimbursements Per Enrollee 



State Market Area Physician and Supplier Services All Medicare Services 





92/91 


93/92 


93/91 


92/91 


93/92 


?3/?l 


AZ 


1.00 


1.00 


1.00 


1.04 


1.02 


1.06 


FL 


0.99 


■ 1.05 


1.03 


1.14 


1.09 


1.25 


LA 


0.98 


1.02 


1.01 


1.13 


1.07 


1.21 


MA 


1.03 


1.04 


1.08 


1.14 


1.09 


1.24 


MI 


1.03 


1.00 


1.04 


1.15 


1.06 


1.22 


MN 


1.00 


1.02 


1.02 


1.09 


1.03 


1.13 


NM 


1.00 


1.01 


1.01 


1.11 


1.06 


1.18 


OH 


l.OI 


0.97 


0.98 


1.11 


1.01 


1.12 


PA 


1.05 


0.99 


1.04 


1.15 


1.04 


1.20 



Source: National Claims History file, 1991-1993. 

* Appendix B contains maps showing the 3-digit ZIP Code areas selected from each state. 



Our advisory committee had major sites in eleven states, including Texas and New York. 
Data problems generally precluded us from analyzing trends in Texas, including Scott & White 
Clinic. We did not include the providers in rural New York (Upper Hudson Primary Care 
Consortium, etc.) in this report because of our emphasis on large, multispecialty groups. Lahey 
Clinic and Fallon Clinic are in Massachusetts. . Although there was only a modest overlap in their 
service areas, we combined the two Massachusetts areas when calculating market level trends. 



24 



for all Medicare services occurredln Florida (25%), Massachusetts (24%), Michigan (22%), Louisiana 
(21%), Pennsylvania (20%), and New Mexico (18%). Lower increases occurred in Arizona (6%), Ohio 
(12%), and Minnesota (13%). 

These differences reveal the problems associated with national average target rates of increase. 
Faced with a national average target rate of increase, a provider in Arizona would have less difficulty 
than a provider in Massachusetts, for example. Therefore, the target should reflect local market 
conditions for the provider, and/or the growth rates in market areas where patients reside. For most 
providers, these two alternatives are virtually the same. However, for other providers this could be a 
worthwhile distinction. For example, some providers are located in tourist or seasonal migration areas 
and serve a significant number of patients who live elsewhere for much of the year. Other providers are 
regional or national medical centers that serve patients from all over the world, and Medicare patients 
from all over the country. 

Observed differences in growth rates between areas probably reflect both random variations, and 
short term or long term systematic factors. The difficulty forecasting short term growth rates for a 
particular area is not limited to HCFA and Medicare. Private and public insurers have to forecast fiiture 
incidence rates of illness, changes in practice patterns, and utilization rates for enrollees. Managed care 
organizations and other providers with a watchful eye on utilization patterns, however, can exert some 
control over patterns of change over time. 

HCFA addresses this issue explicitly in its payment system for HMO risk contractors, the 
AAPCC, which bases rates on the enroUee's county of residence.^ Annual rates of increase for each 
county in the nation are derived fi-om the average national increase times a five-year average of the ratio 
of county to national reimbursement rates. As reimbursement rates in a county show systematic changes 
in relation to national trends, the historical ratio changes, affecting the forecasts for that county. At the 
same time, the five-year moving average smooths out the effects of year-to-year stochastic variations in 
the ratio. This approach involves forecasting error, but over several years the predictions are fairly 
accurate. Difficulties in forecasting short term utilization outcomes also support using longer term, 
cumulative targets (discussed below). 

HCFA could apply the AAPCC approach on a county basis, as it does for capitation rates 
for aged and disabled. Alternatively, HCFA could choose other geographic areas, such as the 
state, which underlies capitaUon payment rates for ESRD beneficiaries. Because HMOs enroll 
Medicare beneficiaries fi-om more than one geographic area (i.e., county), their payments reflect 
the geographic dispersion of enrollees. So, too, HCFA could adjust national projecfions to 
specific provider organizations by applying the data and methods underlying AAPCC rate 
projections, using a set of counties that are relevant to that provider (as with each HMO). There 
are alternatives for defining relevant geographic areas, such as: 



^HCFA sets AAPCC rates at the state level for ESRD beneficiaries. 



25 



• Whole state(s), 

• Counties, 

• Metropolitan Statistical Areas and designated rural areas, and/or ZIP Code areas. 
HCFA has alternatives for deciding which states, counties, etc. to use for a particular group: 

• Within a specified radius of the provider's service delivery sites, 

• Where the provider has service delivery sites, 

• Areas a significant proportion of the provider's Medicare patients reside, or 

• Each area where at least one Medicare patient seen by the provider resided. 

For the empirical results presented below, we used a random sample of beneficiaries drawn from 
the market areas for each provider. The market areas included the three-digit ZIP Code areas in 
which at least 5 percent of Medicare patients resided who were seen by the original or "nucleus" 
site.* On the NCH files, the only geographic identifier that beneficiaries and providers have in 
common is the ZIP Code. 

C. Provider Level Versus Market Level Changes 

In addition to addressing differences in growth rates at the market level, it is necessary to 
investigate whether measures of average resource consumption, specifically RPUPS, are 
sufficiently stable at the provider level. Do the values of RPUPS "jump around" due to random 
factors to such a degree that targets based on a previous year are meaningless? Or, do changes in 
RPUPS bear systematic resemblance to market level changes in reimbursement rates? 

We examined these questions using the provider samples described earlier, and in 
Appendix B. Average market level changes were estimated using the random samples of 
beneficiaries drawn from each of the areas. The resulting rates of change were shown 
previously, in Table 2. We now compare the changes for each provider to the applicable rates of 
change at the market level. 



*We found significant geographic clustering of most patients around the sample 
providers. Although the five-percent rule strictly applies only to the ten (advisory committee) 
sites, the geographic market areas probably are a reasonable approximation of the market areas 
served by the 78 providers. 



26 



Figure 5 shows results for-eomparisons betAveen the average market level changes and 
provider level changes between 1991 and 1992. The origin of the axes represents a perfect 
prediction for changes in the value of a provider's RPUPS based on average changes at the 
market level. The horizontal axis shows deviations from the target for physician and supplier 
services; the vertical axis shows deviations from the target for all Medicare services. Recall that 
these providers were selected nonrandomly, and each typically represents a small percentage of 
beneficiaries in each area. Consequently, the average provider's rate of change need not 
automatically correspond closely to the average market level rate of change. 

There is a significant concentration of providers in the third quadrant, i.e., where the 
actual growth rates at the provider level were less than the average change for the market. The 
second largest concentration is in the first quadrant, reflecting higher rates of change for the 
provider than is average for its market. Perhaps most encouraging is that the large majority of 
providers are huddled within plus or minus 10 percent of the origin on both axes. The providers' 
average unweighted absolute deviation from the relevant target was 4.6 percent for physician and 
supplier reimbursements, and 6.2 percent for total Medicare reimbursements. 

There are at least two points that deviate substantially from the market average rates of 
increase. These are marked Points A and B on Figure 5. These are not among our primary sites, 
and we do not know a great deal about them at this time. However, we do know that Provider A 
saw 63 percent fewer Medicare patients in 1992 as compared to 1991. Similarly, Provider B saw 
58 percent more Medicare patients in 1992, compared to 1991. In contrast, the average absolute 
percentage change in the number of patients seen for the remaining providers was 13.5 percent. 
Removing Providers A and B reduces the unweighted average absolute percentage deviation 
from the relevant target from 4.6 percent to 3.9 percent for physician and supplier 
reimbursements, and from 6.2 percent to 5.4 percent for total Medicare reimbursements. 

Figure 6 shows similar results using values of RPUPS for 1992 predicting 1993. Once 
again, there is a cluster of points deviating from both targets by only single-digit percentages, 
and a smaller group with larger deviations. Several of the "outliers" are marked as Providers C 
through H. Not being any of the providers with whom we have worked, again we know 
comparatively little about them. Only one of them saw more than 4,000 beneficiaries in 1993 
(Provider C), while one saw only about 400 (Provider F), a substantial drop from the number 
seen in 1992. In general, these providers had large changes between the years in the number of 
Medicare patients seen. For example. Provider C served nearly three times the number of 
beneficiaries in 1993, as compared to 1992. 

We suspect that these (and other) providers have undergone significant changes in their 
size and composition over time. The resulting "shock" to the values of RPUPS from unseen 
changes in the providers is less of a policy or data concern at this time, but more likely serves to 
highlight the limitations of observing organizations over time using a single identifier. We 
presume, accordingly, that when physician groups work with HCFA under GVPS, information 
routinely given to HCFA will include changes in their organizational composition. For example. 



Figure 5: Deviations from Targets for 78 Providers in 10 Market Areas: 1991 to 1992 



c 

0) 

E 
<u 
(/) 

u. 
X) 

E 

0) 

cn 

"to 
o 



O) 

I 

H 

E 
o 



in 
c 
g 

• 

TO 
'> 

0) 
Q 

(U 

O) 1 

ni 
■*-' 

c 

(U 

o 

i_ 
0) 

a. 



0.4 



0.3 



0.2 



A 



♦ 



-0.2 



B 



-0.1 



-0.2 



0.1 



0.2 



0.3 



0.4 



Percentage Deviations from Targets for Physician/Supplier Reimbursements 



Source: NCH file. 1991 and 1992. 

Note: Uitlization measures for markets and providers are Reimbursements Per Unique Patient Seen (RPUPS). 



00 

CM 



Figure 6: Deviations from Targets for 78 Providers in 10 Market Areas: 1992 to 1993 




Percentage Deviations from Targets for Physician/Supplier Reimbursements ► 



Source: NCH file. 1992 and 1993. 

Note: Uitlization measures for markets and providers are Reimbursements Per Unique Patient Seen (RPUPS). 



29 



health systems around the nation have been acquiring primary group practices. When this 
happens, such a group would inform HCFA of the change. Services provided through the newly 
acquired primary care practice would be folded into the values of RPUPS for baseline and 
performance periods so that abrupt changes do not invalidate the target. 

Figures 7 and 8 show results for ten of the original physician groups with whom we have 
worked during this study. For these providers, the actual RPUPS in the second year ranges from 
about 4 percent higher than the target to about 7 percent less than the target. From 1991 to 1992, 
five of the ten providers were in the third quadrant; from 1992 to 1993, seven of the ten providers 
were in the third quadrant. Of the three groups in other quadrants (Groups M, N and O), two 
were also outside the third quadrant in Figure 7 (M and O). 

D. Cumulative Versus Year-to-Year Targets 

MVPS has been implemented on a year-to-year basis. This approach is relatively 
"forgiving" in that actuarial projections underlying the expenditure targets take into account the 
most recent utilization patterns. There are proposals to designate a base year and make future 
targets (and penalties) cumulative. Long run cost control may be achieved more reliably if target 
rates of increase in Medicare expenditures are set with respect to a specific baseline year. For 
example, the target for Year 2 could be specified as 5 percent above expenditure levels in Year 1; 
the target for Year 3 could be 1 percent above expenditure levels in Year 1 , without regard to 
actual expenditures in Year 2. 

For GVPS, each approach has advantages: cumulative targets could better reward 
successful groups, while year-to-year targets may encourage more groups to "give it another try" 
each year. The ability of physician groups to alter utilization patterns may vary from one year to 
the next. Sizable efficiencies in one year may be followed by less relative change in the next 
year. If the group was still ahead of the average, its relative efficiency still could be rewarded 
through cumulative targets. If targets are always based on the most recent data, then the target 
each year would be reduced in accordance with the improved efficiency. Consequently, rewards 
would be harder to generate. Groups might have to "pace themselves" with incremental 
improverhents each year in order to achieve the same rewards for cumulative gains that would 
occur using cumulative targets. 

On the other hand, participating groups that fail to meet targets in the early years would 
find it more difficult to "get below" the cumulative target. This has two aspects. First, if a group 
really tries to manage services but has higher than average growth rates, it may conclude that 
meeting future cumulative targets is unlikely. Second, HCFA could track cumulative savings (or 
deficits) for the group and not pay rewards unless or until cumulative savings are positive. If a 
group accumulates a deficit, it may conclude that the present value of future reward payments is 
very low. Furthermore, if special penalties await groups that fail to meet their targets (discussed 
below), participation under GVPS could be brief. 



o 

CO 



Figure 7: Deviations from Targets for 10 Selected Providers: 1991 to 1992 

0^6- 



c 

0) 

E 
{2 



^ -0 
i2 



«» 

c 
g 

n 
■> 

0) 

Q 

0) 



0.04 



0.02 



M 
♦ 



^7 



a. 



-0.06 



-0.05 -0,04 



-0.03 



-0.02 
J 



K* 



N 



-0.01 



51 



-0.02 



-0.04 



-0.06 



-OtOS- 



Percentage Deviations from Targets for Physician/Supplier Reimbursements 



Source: NCH file, 1991 and 1992. 



Figure 8: Deviations from Targets for 10 Selected Providers: 1992 to 1993 



-0t04- 



c 
E 

0) 

!2 

3 

E 
■<u 
cd 

ro 
o 



0) 
(0 



E 
o 

c 
o 

ra 

■>■ 

0) 
T3 

0) 
O) 

ro 
c 

(U 

ij 

CL 



08 



-0.07 



K 



-0.06 



J 
♦ 



P 
♦ 



-0.05 



-0.04 



1/r 



0.02 



O 



-0.03 -0.02 ♦-O.O! 

R 



-0.02 



-0.04 



-0.06 



-0.08 



-0.1 



-0r45- 



0.01 



N 



Percentage Deviations from Targets for Physician/Supplier Reimbursements — ^ 



Source: NCH file, 1992 and 1993. 



32 



Figure 9 shows the results'of predicting RPUPS for 1993 by multiplying RPUPS for 1991 
times cimiulative rates of increase for each respective market area. Since these are the same 
providers shown earlier, the deviations are additive. If providers' deviations from market 
averages continued in the same direction (higher or lower than the market), the data points would 
tend to move away from the origin (the market prediction). If deviations were more random and 
therefore canceled out over time, the points would be closer to the origin after two years than the 
one-year comparisons. In general, the cumulative deviations tended to be larger (8.1 percent for 
physician/supplier services, and 9.5 percent for all Medicare services) than for the one-year 
targets (4.6 percent for physician/supplier services, and 6.2 percent for all Medicare services). 
The aforementioned specific points that changed significantly during this period (i.e., A through 
H) are shown again in Figure 9. 

Figure 10 shows the cumulative targets for the 10 selected physician groups. Eight out of 
the 10 providers were in the third quadrant, indicating lower rates of increase compared to the 
average market changes, for both physician/supplier services and all Medicare services. Only 
Providers M and O were above their respective market level rates of increase for either scope of 
service. Most of these multispecialty group practices appear to be manifesting lower rates of 
growth than is average for their respective market areas. 

The approach of a cumulative target may be workable, given the reasonable stability of 
provider level measures of RPUPS relative to the market level random samples. A cumulative 
target embodies more frilly the principle behind payment system reform, namely to set in place 
incentive systems that work toward long run cost containment. Perhaps HCFA could introduce 
ftirther refinements of this approach, such as fixing any errors that may have resulted in the 
process of setting targets. Examples would be to adjust for errors made in forecasting exogenous 
factors such as the rate of inflation or growth in gross domestic product, if these are factors 
HCFA would use in setting targets. 

Another issue is selecting the base period from which to make projections. For a policy 
that begins soon, the latest data available would be for 1993 or 1994. Some groups may prefer to 
use an earlier year because they have implemented a number of managed care systems in recent 
years and would prefer credit for these achievements. The earliest HCFA probably would 
consider is 1992, because so many changes have taken place in reimbursement policy (e.g., the 
MFS) since 1991. A related issue is whether and when to "rebase," i.e., to choose another actual 
year of data from which to make further cumulative targets. We recommend that HCFA wait 
before deciding when to rebase the GVPS, and simultaneously consider ways to measure cross- 
sectional differences in RPUPS. Including methods for recognizing differences in efficiency 
within a given time period could help avoid the problem of lowering targets for physician groups 
essentially because of their successes to date. 



en 



tn 
c 

0) 

E 

0) 
X> 

E 
<u 
oc 

3 
o 



0) 
O) 

.ro 



E 
o 

C 

g 
*■*-« 

(0 

■> 

lU 

Q 

0) 

cn 
ni 

•4-* 

c 

0) 

t> 

0) 
Q. 



Figure 9: Deviations form Targets for 78 Providers in 10 Market Areas: 1991 to 1993 

0:6- 



-0.4 



B 



0.4 



0.2 



-0.2 



H 



♦ 

♦ ♦ 



— - ♦ ♦-"o^ 



-0.2 



-0.4 



0.2 



0.4 



Percentage Deviations from Targets for Physician/Supplier Reimbursements ^ 

Source: NCH file, 1991 and 1993. 



14 



Figure 10: Deviations from Targets for 10 Selected Providers: 1991 to 1993 



-0.12 



-0.1 



-0.08 



K 



-0.06 



R 



-0.04 



-0.02 



-0.Q5- 



N 

-0.05* 



-0.1 



-0.15 



M 



Percentage Deviations from Targets for Physician/Supplier Reimbursements ^ 



Source: NCH file. 1991 and 1993. 



35 



VIII. Rewards and Penalties ' 

A. Policy Questions and Objectives 

The potential economic consequences of GVPS for physician groups and for HCFA are 
discussed in this section. The factors that determine the economic results of GVPS have to do 
with the incentives and financial transactions related to rewards and penalties. 

The MVPS system links physicians' fees to national aggregate expenditures in order to 
recoup excessive payments. In that context, penalties are instrumental in achieving budgetary 
goals. As alternatives to the current national approach are considered, the role of rewards and 
penalties needs to be assessed carefully. In particular, 

► Should voluntary participants face potential penalties that exceed those imposed on other 
physicians? 

► How could the amounts of the rewards or penalties be determined? 

► What are the sources of funding for rewards, assuming budget neutrality? 

► What mechanisms could be used to administer rewards and/or penalties? 

The policies chosen in response to these questions presvraiably ought to reflect the 
assumptions behind, and the objectives for, GVPS. Our discussion assumes that HCFA would 
desire to: 

»• Encourage participation among qualified providers if broad participation would achieve 
the largest savings for Medicare, 

*■ Encourage and reward successful improvements in the efficiency of practice — an 
ingredient normally missing in the fee-for-service sector, and possibly to 

► Encourage beneficiaries to receive care fi-om the most efficient providers. 

The remainder of this section addresses the policy issues involving rewards and penalties for 
groups under GVPS specifically, and under MVPS generally. 

B. Options for Rewards and Penalties 

The table below conveys four general scenarios that could occur if a GVPS option were 
available. The scenarios are defined in terms of whether a hypothetical participating group 
and/or physicians nationally meet their designated expenditure targets. 



36 



GVPS Met? 


National MVPS Met? 


Yes 


No 


Yes 


Both targets met 


Only group target met 


No 


Only national target met 


Neither target met 



Without GVPS, there are two observed scenarios: the nation does or does not meet the 
performance standard(s). Although individual groups may be improving their efficiency relative 
to the average even without explicit incentives, these differentials are unobserved and 
unrecognized. In addition, the adoption of GVPS could help to foster improvements in 
efficiency among some or all participants, which would be observable as the four scenarios 
portrayed above. There also is the issue of how much the actual outcomes differed from the 
respective targets, which can be included in deciding the amount of the rewards or penalties. 

In principle, it may be advantageous to differentiate all physician groups, and assess their 
volume performance separately. In jjractice, there are administrative and other factors that might 
make that goal unattainable. HCFA may wish to assess the volume performance of qualified, 
selected groups separately, and the remaining providers together as an undifferentiated group. It 
also may be desirable, in principle, to reward or penalize each group that is differentiated under 
its own performance standard, including the residual physician population. However, at this 
point, Congress has called for potential differentiation on a voluntary basis. Among our advisory 
committee, we found that models with penalties concentrated on groups that fail to meet their 
target might be acceptable to some physician organizations, but not to others. We offer model 
variations with and without concentrated penalties for volunteering organizations. 

The options associated with successful performance of groups under GVPS include to: 

► Avoid or refund blanket penalties that occur under MVPS, without giving extra 
rewards. Groups could be exempted from any general penalties imposed on physicians' 
fees in the event of unfavorable national performance. 

► Avoid or refund blanket penalties, and give extra rewards. Groups could earn 
rewards in addition to avoiding penalties. This could provide extra incentives for 
participation, and reward efficiency for services other than physician services (e.g., 
hospital savings). 

► Give extra rewards, without avoiding or refunding blanket penalties. Groups could 
be rewarded for their observed successes, but still be subjected to penahies that apply to 
all physicians in the nation. 



37 



The options associated with unsuccessful performance of groups under GVPS include to: 

*■ Limit penalties to any changes in the national conversion factors. In the event a 
group fails to meet its target, the group could receive the same penalties (if any) as 
physicians in the national pool. For example, the single national conversion factor(s) 
would apply to the group, and no offsetting financial reward would be forthcoming. This 
is equivalent to a "hold harmless" provision for GVPS. 

»• Impose concentrated penalties. An alternative would be to impose penalties on 
participating groups that fail to meet their own targets. This could be justified as an 
additional incentive for efficiency. Alternatively, this may be viewed as an equitable 
punishment for allov^ng Medicare expenditures to rise faster than the Federal 
government's specified rate. A third reason may be perceived equity fi-om symmetry of 
upside and downside risks. Fourthly, HCFA may wish to establish each group as a 
separate population, between which there is no subsidization of penalty amounts. Lastly, 
concentrated penalties may deter participation by groups that are not likely to succeed, 
saving administrative costs. This would function similar to eligibility criteria, i.e., to 
limit participation to the groups most likely to succeed. 

Refunding national fee penalties to groups that met their own GVPS would help to 
achieve equity in payment. Rewards and/or penalties could be implemented in order to achieve 
equity in payment, to help induce efficiencies, or to affect participation patterns among groups. 
Sharing the program savings with groups that beat their own GVPS would instill fundamental 
changes in the incentive structure that exists under fee-for-service. In either case, implementing 
a GVPS model that allows for potential rewards but not concentrated penalties may stimulate 
many physician organizations to volunteer for GVPS. In that context, non-participation may 
result from factors such as: 

► Insufficient interest (e.g., proportionally few Medicare patients), 

► Failure to meet other criteria for qualification (e.g., narrow scope of practice), or 

► No desire to encourage scrutiny of utilization patterns. 

There may be a connection between the attractiveness of concentrated penalties and the 
method for setting targets (i.e., cumulative versus rebasing each year), and the method for 
estimating Medicare Savings (i.e., cumulative and/or yearly). Rebasing a group's target each 
year based on actual utilization — in the absence of potential concentrated penalties — might lead 
to incentives for the group to inflate its base in some years, alternating with years of reaping 
rewards for artificial savings. Penalties could reduce the benefit of this ploy. Similarly, 
cumulative targets could create a natural penalty situation when a group inflated utilization rates 



38 



unless a provider was able to chum patients "at will" and yet demonstrate long run relative 
efficiency. The potential benefit of gaming could be reduced even ftirther if HCFA paid rewards 
to a provider only if cumulative Medicare Savings were positive. 

If the GVPS option is adopted by greater numbers of physician organizations, the 
imdifferentiated physician population would become smaller. Over time, the existing GVPS 
groups could grow in size as well. HCFA may choose later to end "hold harmless" provisions 
for groups operating under GVPS (i.e., add concentrated penalties). Criteria could include the 
number of beneficiaries seen or the nimiber of years operating imder GVPS. In the mean time, 
some groups may choose to face potential penalties in exchange for a higher share of the savings 
generated under GVPS. 

C. Reward and Penalty Formulas 

Under GVPS, providers would continue to be paid according to HCFA's standard 
payment policies for all types and places of service. There may be additional economic 
consequences for providers, depending on their own performance, the national performance 
under MVPS, and other potential long-term cost control policies. This section discusses methods 
for determining the value of savings, rewards and penalties under GVPS. 

Success iinder GVPS means not exceeding the target rate of increase specified by HCFA, 
for this satisfies the budgetary goals of the system. Under MVPS, meeting the target avoids 
penalties to fee updates. Similarly, a starting point for GVPS could be the avoidance of financial 
penalties for groups that meet their own performance standard. The economic value of that 
scenario would be the relative value units of the physician and supplier services delivered by the 
group, multiplied by the difference in the conversion factor caused by the national penalty. A 
lump sum payment to the group could offset the lost revenue due to lower Medicare payment 
levels — i.e., a refund of the penalty. 

The proposed GVPS models go beyond making technical corrections to MVPS such as 
refunding the actuarial value of the penalty. Instead, we offer policy options that would revamp 
the incentive structure created by the payment system, and give providers tangible motivation to 
manage patient care more vigorously. GVPS could evoke a mind set similar to capitation. In 
addition, GVPS could give providers a return on the investments that go hand-in-hand with 
managing care. Providers carmot make adequate investments based on an expected return 
equivalent to the reductions in updates to physicians' fees. 

The fee-for-service payment system gives more revenue and higher net income to 
providers that deliver higher volumes and intensities of services. A provider's net income is 
based on the difference between reimbursements for services and the cost of providing the 
services. Capitation changes this by paying providers the full expected reimbursement level even 
for services that are not delivered. Thus, for ayoidable services the provider's net income 



39 



increases by the gross reimbursement amount. Net income increases by reducing utilization 
rates. 

Formulas for rewards and penalties under GVPS could vary the incentives and economic 
consequences of providers' performance between these extremes. The basic form of payment 
still is fee-for-service, and rewards and penalties could be very small in comparison. This would 
represent perhaps a modest departure from the status quo. Alternatively, the rewards and 
penalties could be potentially large and structured to more closely approximate capitation. 

Methods for determining the amount of rewards and penalties could include these four 
variables: 

► Refund. A provider can lose revenue because HCFA has imposed blanket penalties on 
prevailing payment rates. Penalties under MVPS force all physicians to "pay for" 
excessive national Medicare expenditure levels. Demonstrating relative efficiency and 
meeting HCFA's allowable growth rate under GVPS could allow a group to avoid the 
economic consequences of those penalties. HCFA could, in effect, refund the group's 
penalty payments. 

Refund = (Sum of Relative Value Units for Physician and Supplier Services) x 
(Difference in Conversion Factor(s) due to Penalties) 

► Medicare Savings. This is the amount of savings estimated from the observed difference 
between actual and expected values of RPUPS for the provider. The savings are the 
difference between the reimbursement rates, multiplied by the number of beneficiaries 
seen in the performance year. 

Medicare Savings y = (Target' y - RPUPS y) ^ N y 

where Y is any year a group's performance is evaluated, 

Target' is the predicted RPUPS y, adjusted for differences in the provider's case 
mix between the base year and the performance year, and 

N is the number of Medicare beneficiaries who received physician services from 
the group (i.e., the denominator in RPUPS y). 

Patient Capture Ratio (PCR). This is the proportion of reimbursements to all 
providers that went to the physician group operating under GVPS, for patients included in 
the measure of RPUPS. The PCR may be defined only in terms of physician and supplier 
services, or may be expanded to include all Medicare-covered services, in accordance 
with the scope of services included in RPUPS. Accordingly, a health system that owns a 



40 



hospital might include payments to its hospital in the numerator (and the denominator) of 
the PCR.^ 

*■ Sharing Rate. HCFA could choose a particular Sharing Rate, either alone or in 

conjunction with the PCR. Using a Sharing Rate alone would recognize the equivalent 
value to Medicare of savings from intemal efficiencies by the provider (i.e., reducing 
utilization within its own system), or external efficiencies (e.g., reducing utilization rates 
of an unaffiliated hospital). Using a Sharing Rate with the PCR could approximate more 
closely a provider's actual forgone net income under GVPS, rather than forgone gross 
revenues. 

Thus, we specify the potential formula:* 

Reward = Refund + (Medicare Savings x Patient Capture Ratio x Sharing Rate) 

GVPS would induce efficient practices if the reward system was like capitation, in which 
reducing utilization increases net income. The formula for sharing the savings could offset 
reductions in revenues by giving the group a sufficient proportion of total observed savings. One 
option would be to: 

*■ Pay to the physician group the amount equal to the Medicare Savings times the Patient 
Capture Ratio. (This sets the Sharing Rate to 1 .0). 

»■ HCFA would keep the remainder of the savings, i.e., Medicare Savings times the inverse 
of the PCR. 

Reward = Medicare Savings x Patient Capture Ratio x (1.0) 

This approach would return to the physician group all revenues forgone under fee-for-service 
resulting from utilization rates that were lower than actuarial projections would indicate for 
"average physicians" seeing that population of patients. Physician groups would be treated as 
"virtual HMOs" paid a "virtual capitation" that is prorated according to the proportion of the 



''We are exploring, conceptually and empirically, the implications of expanding the PCR 
to include other services. 

* Refund refers to the actuarial value to the group of the national fee penalties. This could be 
added to any of the variations on reward formulas discussed here. 



41 



patient population captured by the' group. HCFA would keep the remainder of the savings.' The 
physician group would not have expected that revenue in any event. 

Other values could be given to the Sharing Rate. For example, HCFA could: 

*- Reduce the Sharing Rate to some fraction. 

Reward = Medicare Savings x Patient Capture Ratio x (0.5) 

This would move away from the incentives closely resembling capitation and toward a system of 
compensating groups for lower net income. Another option would be to: 

Pay to the physician group a specified or negotiated Sharing Rate, without regard to the 
PCR. 

This amount could be set rather arbitrarily, such as one-half, and applied to all physician groups. 

Reward = Medicare Savings x (0.5) 

We prefer the option that bases payments to the physician group on the Patient Capture 
Ratio. One very nice property of this approach is that using the PCR to gauge rewards would 
prevent HCFA from overpaying rewards when beneficiaries are seen by more than one provider 
under GVPS. Thus, any number of groups could be operating under GVPS, with many seeing 
the same patients, and total reward payments would not be excessive due to "double counting." 

However, there are issues that need to be considered from the provider's perspective 
about using the PCR. Groups that helped to avoid outside utilization, or that moved outside 
utilization to inside, would increase their PCR and their share of the savings. This seems to be 
an acceptable arrangement. However, reducing inside utilization rates (enhanced internal 
efficiency) would increase the Medicare Savings but decrease the PCR. To address this, HCFA 
could place a floor on the PCR, for example, its value in the base year. Similarly, HCFA could 
use the group's PCR in a certain year (e.g., the base year to start) in the reward formula for a few 
years (e.g., three to five years), then reevaluate or rebase the parameters of the formula using an 
updated value of the PCR. 

If the policy called for concentrated penalties for a group that exceeded its performance 
standard, HCFA could be take a similar approach. 



^ A hypothetical physician group with an extremely high Patient Capture Ratio would 
resemble an HMO. Perhaps HCFA could limit the share paid to a group to 95 percent, thus 
emulating the implied savings to HCFA under risk contracting. 



42 



Penalty - Medicare Losses x Patient Capture Ratio x Sharing Rate , where 
Medicare Losses = (RPUPS y - Target'y) ^ N y , 

where Y is any year a group's performance is evaluated, 

Target' is the predicted RPUPS y, adjusted for differences in the provider's case 
mix between the base year and the performance year, and 

N is the number of Medicare beneficiaries who received physician services from 
the group (i.e., the denominator in RPUPS y). 

Here again, HCFA must choose values for those variables that reflect the appropriate incentives 
and risks for participating groups. The PCR represents a gauge on what proportion of 
reimbursements for the patient population went to the group, hence its direct benefit fi-om the 
excessive utilization. For this reason, we believe that PCR would be a usefiil component of the 
penalty formula. It would seem reasonable to limit penalties through a fractional Sharing Rate 
since groups would not have the authority over utilization that comes with enrollment and "lock- 
in."'° 

It may be reasonable for HCFA to specify distinct combinations of reward and penalty 
formulas. For example, a model without any penalty might have a reward formula with a lower 
Sharing Rate than a model with a penalty, as was illustrated in Table 1 . Similarly, larger 
potential rewards could be provided in conjunction with larger potential penalties. 

In addition to measuring Medicare savings each year, HCFA may choose to evaluate 
Cumulative Medicare Savings attributable to a group's performance. In that context, each year 
the group will make a positive or negative increment to its cumulative savings. Whether a group 
has accumulated a surplus or a deficit may affect the economic consequences of Medicare 
Savings in any given year, in terms of rewards or penalties. 

D'. Transaction Mechanisms 

Under MVPS, HCFA imposes penalties by reducing the conversion factors (CF) that 
determines all physicians' fees. This is administratively simple because there is only one 



'°It would be appealing to set targets for a group's own Medicare revenues, for example 
to determine whether a provider benefited from excessive overall utilization. Although this type 
of question may be worthwhile in the context of an evaluation, it may be difficult to project 
expected reimbursement rates for idiosyncratic service mixes. This is discussed further in 
Appendix C. 



43 



conversion factor per category of services — surgical, non-surgical and primary care. A 
disadvantage is that penalties are deferred two years into the fiiture. 

The mechanism of adjusting CFs could be applied to GVPS, at least for physician and 
supplier services. This would result in different CFs for each participating group — many more 
than the one or two national CFs, but not nearly as many distinct prices as were maintained by 
carriers under the usual, customary, and reasonable (UCR) system that preceded the Medicare 
Fee Schedule. In effect, rewarding providers through differential conversion factors would 
distort the relative values across services. Furthermore, beneficiaries might be given a 
disincentive to use the group because higher CFs presumably would lead to higher copayments. 
Given those situations, multiple conversion factors may turn out to be complicated and 
undesirable. 

Alternatively, a reward could be accomplished through a lump sum payment to the 
physician group. After the end of a successftil year, and after the financial calculations could be 
made, the physician group would receive a single payment equal to the value of the reward. 
Disbursements to physicians within the organization would be an internal matter, although data 
from HCFA regarding internal and external utilization perhaps could be used to guide 
decision-making. For example, the group could analyze the causes for savings (by place of 
service, physician specialty, etc.). 

For models with concentrated penalties, a mechanism for achieving those transactions 
also would be needed. Using conversion factors is not desirable, as discussed above. Because 
the groups presumably are ongoing businesses with continued involvement in treating Medicare 
patients, HCFA could administer the penalty by deducting the value of the penalty from future 
payments to the group. For example, if a penalty of $50,000 is owed by the group, payments to 
the group in the following year would be reduced by that amount. This would be a simple 
deduction and would not affect conversion factors or relative values for individual services. 

In addition to the specific transactions with the groups, there may be other economic 
consequences siffecting physicians generally. Decisions about sources of funding for rewards, 
and the amount of savings kept by Medicare, may affect the conversion factors for all physicians. 
A problem could arise under GVPS if it is necessary to know what would have happened in 
terms of RPUPS values for the group in the absence of GVPS. In the reward and penalty 
formulas discussed above, the Medicare Savings (and Losses) were defined as the difference 
between actual and expected RPUPS. Granted, the savings may be real and even attributable to 
the group, but some of those savings may have occurred in the absence of GVPS. On the other 
hand, the observed market rate of increase may have been an underestimate of increases that 
would have occurred in RPUPS for a given provider. In other words, GVPS may have induced 
even larger savings than are apparent. 

Rewards to groups for savings that would have occurred anyway can be justified on 
equity grounds — at the very least rewards up to the actuarial value of national fee penalties. 



44 

Rewards to groups for savings that would not have occurred with GVPS are appropriate 
incentive payments. However, rewards to groups for savings that would have occurred anyway, 
although difficult or impossible to estimate, Would increase Medicare costs relative to the 
program without GVPS. Therefore, HCFA needs to consider the policy with respect to its 
implications for physicians generally: It is likely the conversion factors for all physicians would 
need to be reduced by some amount to recoup some or all of the value of rewards to successful 
groups. The argument may apply to penalties as well: Special penalties imposed on groups for 
relatively high growth rates could offset a portion of excess payments that would otherwise be 
financed by broad-based penalties. 

Whether a reward is for equity or to induce efficiency, HCFA could adjust its accounts 
under MVPS when determining the national conversion factors. The adjustment could: 

*■ Add back the Medicare Savings amount to the national expenditure totals. This 
would effectively remove the successful GVPS groups from the evaluation of national 
performance, because each group would be imputed an average rate of increase. Thus, 
HCFA could assess the collective performance of the remaining, undifferentiated 
physician population. 

*■ Add back just the amount of the Rewards to the national expenditure totals. Under 
this approach, the nation would still get some credit for the successful performance of 
groups. However, HCFA would finance the actual rewards through lower fees to 
physicians generally. 

Another type of difficulty arises because RPUPS may be defined in terms of all Medicare 
services, not just physician and supplier services. For Part A and other Part B savings, HCFA 
needs another mechanism to finance rewards and to focus penalties more on the inefficient 
providers. Currently, there is no fi-amework corresponding to MVPS for facility and other non- 
MVPS services. The analog to MVPS fee updates would be PPS updates, or allowable cost 
increases for cost-reporting facilities. As HCFA currently considers many factors when setting 
updates for these costs, the reward system under GVPS could become another relevant factor in 
those determinations. 

E. Possible Implications 

The concept of a reward is introduced in GVPS for two reasons. First, it permits a 
significant reallignment of incentives to the provider that is more akin to managed care than fee- 
for-service. Second, it allows for a return on a provider's investments in managed care systems, 
which might be more costly than the discounted present value of fiiture fee penalties, but very 
worthwhile in relation to potential Medicare cost savings. 

Under MVPS, physicians generally will be subjected to penalties to the extent that 
national Medicare physician expenditures fail to meet target levels. Physician groups operating 



45 



under GVPS that improve their own efficiency also reduce total Medicare expenditiu-es. 
Resulting decreases in volume and/or intensity of services are financially beneficial to 
Medicare — even if the changes are not sufficient to have overall group performance meet target 
levels under GVPS. 

There are several basic objectives we set out to meet in developing GVPS options, 
including to further equity and efficiency through the payment system and to assure budget 
neutrality for Medicare. If HCFA pays physician groups under GVPS according to their 
estimated savings, other physicians are likely to be made worse off for reasons that are consistent 
with the attractiveness of GVPS. Specifically, providers that show lower rates of increase can be 
better off through GVPS. At the same time, the average performance of remaining physicians 
will be worse because the more efficient providers will be underepresented in the remaining 
physician pool. 

The structure of incentives under GVPS can have implications for the degree of income 
transfer among physician groups. Larger reward payments would lead to lower revenues for 
other providers. Model parameters, such as the definition of the PGR, can affect incentives and 
economic consequences for providers. For example, HCFA will need to determine the 
conditions under which hospital services can be included in the PGR. Including hospital services 
within the PGR for a physician group that owns a hospital may bolster greatly the incentives to 
reduce admission rates. However, a physician group that does not own a hospital, but reduces 
admission rates, would not necessarily be adequately rewarded. HCFA may allow 
circumstances, other than ownership, in which the PGR would include hospital and other 
services. Alternatively, HCFA could adjust the Sharing Rate to magnify the incentives for 
efficiency." 

IX. Eligibility Criteria for GVPS 

A. Linking to Purposes 

Deciding upon criteria for participation under GVPS requires consideration of other 
aspects of GVPS, including the goals of the system, the scope of services involved, and the 
reward and penalty structure. 

*■ Equity in payment. GVPS could be used to improve equity by rewarding efficient 

practices and focusing penalties on inefficient practices. Assuming the more equity the 
better, HCFA could aim to disaggregate the physician populafion as much as is feasible. 
Furthermore, HCFA may wish to consider mandatory versions of GVPS that would 
further sharpen the reward and penalty structure associated with Medicare payments. 



"Values for the Sharing Rate in excess of 1.0 would tend to override limitations imposed 
by the PGR in circumstances such as these. 



46 



► Pervasive incentives for efficiency. In general we could expect greater total savings as 
incentives for efficiency are applied more universally. With more providers participating, 
greater proportions of the Medicare population would come into utilization measures of 
groups under GVPS and the aggregate "patient capture ratio" would increase. Aiming for 
maximum penetration of GVPS might involve establishing only minimal eligibility 
criteria. HCFA could limit criteria to those required for validity of the models. 

► Broad-based managed care initiatives. Alternatively, HCFA could use GVPS 
explicitly to build up a network of providers that have the organizational capability to 
manage total patient care. Providers differ in their ability to manage care. These 
differences can be associated with a number of organizational and experiential 
characteristics that HCFA could use to select as eligibility criteria. 

HCFA may use GVPS as a step toward managed care and preferred provider 
arrangements. Accordingly, criteria for selecting qualified groups may be more substantial right 
from the outset of GVPS. These could include assurances of clinical quality and patient 
satisfaction, for example. Furthermore, if financial penalties are involved, the group may need to 
demonstrate adequate financial standing. HCFA already has developed substantial criteria for 
monitoring managed care organizations, as described in its Contractor Performance Monitoring 
System. 

B. Basic Eligibility Criteria 

There are certain eligibility criteria that may be necessary for the utilization measures and 
performance standards to be valid and stable. These can relate to characteristics of the group 
itself, or to aspects of the patient population seen by the group. These include: 

»■ Formal definition of the group. We have referred to "groups" participating under 

GVPS, and to integrated health care delivery systems. Existing physician group practices 
and health systems may be willing and able to elect separate performance standards. 

However, there may be other arrangements among providers that come about in order to 
participate under GVPS. For example, several physician groups may apply to operate 
under a separate GVPS that is based on their combined experiences. HCFA will need to 
formalize agreements between separate organizations that are named to operate together 
under GVPS. The group practices, solo physicians, facilities, and other providers whose 
Medicare Provider ID Numbers are to be included in the definition of the group need to 
formally agree to this arrangement. Any formal relationships will have to be clarified and 
deemed legally acceptable. For example, there may be legal restrictions on referring 
patients for services when the referring physician has ownership interests. 



47 



>■ Specifications for internal sharing of rewards and penalties. In addition, the 

organization(s) need to have internal arrangements for accepting and disbursing lump 
sum reward payments from HCFA. Similarly, written contracts between HCFA and the 
organization(s) would need to specify methods for the group to settle penalty situations. 
Even if HCFA "receives" the penalty payments by withholding of fiiture reimbursements 
to all providers in the group, there may be a need for guidance about the distribution of 
those withholdings. 

For example, a GVPS group may be comprised of four physician group practices. 
Penalties (and rewards) may be distributed in proportion to their Medicare revenues in the 
performance year. However, if one of the groups gets much bigger in the subsequent year 
it may bear a disproportionate burden if HCFA withholds reimbursements from all the 
providers until the penalty is paid. 

*■ Size. There are likely to be dimensions of size on which HCFA must specify a minimum 
value for a group to participate. For example, HCFA could not set performance standards 
for a solo physician who sees between one and ten Medicare patients per year. Average 
resource consumption for the patients in one year would not be expected to form a 
reliable baseline and performance standard for that physician's future patients. 

That example illustrates two size dimensions that are likely to be important criteria for 
eligibility: the number of Medicare patients seen and the number of physicians in the 
group. There may need to be a large number of patients included in the measure of 
average resource consumption (i.e., RPUPS) because of the relatively large variance 
associated with reimbursement amounts at the individual level. There may need to be a 
minimum number of physicians in the group so that arrivals and departures or shifts in 
clinical emphasis do not inordinately affect the group's case mix. 

Size may be important for a somewhat different reason. HCFA could decide to limit the 
number of groups operating under GVPS for administrative reasons. For each GVPS 
group, HCFA would have to retrieve relevant data elements and calculate parameters 
siich as the Medicare Savings and the Patient Capture Ratio. Although there is a 
correlation between the size of the group and the number of data elements, it may be 
practical to limit the steps to larger groups. This does not necessarily preclude 
participation by any provider, but implies that interested providers may need to pool their 
experiences into larger groups in order to "make the cut." 

If HCFA limited eligibility criteria to these basic requirements, an enormous variety of 
provider groups may be eligible to participate. This is desirable to the extent that HCFA would 
like to have maximum application and maximum total effect of GVPS. In addition, the basic 
criteria could allow expensive single specialty providers to participate, even those without a 



48 



primary care emphasis.'^ This als5 may be desirable since large portions of Medicare 
reimbursements are for specialty services, the providers of which could manage their own 
services. Creating incentives for a wide range of providers to manage their own practices could 
be a positive result of GVPS. 

C. Criteria for Managed Care and Preferred Provider Initiatives 

A different tact for HCFA could be to identify types of organizations that could manage, 
or even provide, a full range of services. The utilization measures and performance standards 
described in this report are based on all reimbursements for patients, including for services by 
other providers.'-' HCFA may undertake a definitive assessment of a group's capacity to manage 
the full range of Medicare-covered services. This may require several layers of criteria. 

The suggested criteria presented below are based on the assumption that a very large 
number of provider groups will want to participate in the program. On the one hand, HCFA may 
want to develop fairly restrictive criteria at the outset so as to best assure the success of the 
program and to provide the necessary administrative oversight. On the other hand, Medicare 
may not be best served by requiring groups to meet stringent selection criteria at the outset, and 
thus excluding potentially successful groups from applying for GVPS. In this situation, criteria 
should be viewed as recommendations, deviances firom which could be justified by groups and 
feasibility for participation evaluated by HCFA. 

With the growth of managed health care plans, large multispecialty medical groups have 
had to compete on price with other providers for contracts. The contracts establish the prices that 
the plans will pay for provision of comprehensive services, and the requisite assurances about the 
quality of the services. While the practices are given varied economic incentives (withholds, 
capitation payments or bonuses) to operate efficiently, these providers are selected by managed 
care plans based on relatively standard criteria and procedures. HCFA already contracts wdth 
HMOs and other managed care organizations and has determined a number of criteria for 
assuring quality of care. These principles, and perhaps criteria used by other payers, could be 
adopted by HCfA to help regulate delivery patterns and utilization management under GVPS. 



The more specialized or abnormal is the mix of services in a group's utilization 
measure, the more necessary it may be for HCFA to adjust projected growth rates. (See 
Appendix C, Section G). 

We gave consideration to measuring average resource consumption based on 
reimbursements to the provider only. This definition was rejected for three reasons: too many 
threats to the validity of performance standards (i.e., changes in a provider's specialty 
composition or scope of service); the difficulties measuring actual savings to Medicare; and the 
willingness of providers in our advisory committee to understand and accept models with 
reimbursements to all providers. 



49 



Even though groups will have incentives to manage the services used by their Medicare 
patients, GVPS models do not necessarily require beneficiaries to make any formal commitments 
to the groups as their patient care managers. The groups may be expected to institute procedures 
and practices that enhance consumer satisfaction and loyalty because beneficiaries can change 
providers. However, an argument for establishing more elaborate eligibility criteria, and for 
monitoring patient satisfaction, etc., is to avoid any inappropriate responses to the economic 
incentives. For example, a provider may restrict the beneficiaries' freedom-of-choice to utilize 
providers outside the group practice, e.g., through slow referrals. Therefore, beneficiaries using a 
group operating under GVPS could be informed of the special arrangements with HCFA. 

► Composition of the group. HCFA may wish to limit the groups selected to those with 
multiple specialties or primary care gatekeepers. Appropriate scope could include 
primary care physicians, general surgeons, internists, and specialists in such areas as 
cardiology, oncology and neurology. By reducing the need to refer patients outside their 
own system, such groups might lose a smaller proportion of their capacity to efficiently 
manage their patients' clinical needs. 

In addition to integrated multispecialty groups, primary care-oriented physician groups, 
such as family practitioners and/or internists, could function effectively as patient 
managers. They may have developed patterns of referrals to specialists whom they 
expect to be efficient. They may eventually, under the aegis of GVPS, develop more 
formal alliances with preferred providers that would further enhance the delivery of cost- 
effective care. 

»- Accessibility standards. Because managing care can require swift responses, HCFA 
may wish to assure that Medicare beneficiaries have access to medical care in the group 
on a continuous basis. Accordingly groups could be required to demonstrate that triage 
systems are in place so that urgent visits (as well as symptomatic visits) may be 
scheduled on a timely basis. Medicare beneficiaries should be able to count on 24-hour 
access to the group practice for emergencies. Groups applying to participate in GVPS 
should be positioned to provide appropriate telephone coverage when physician offices 
are closed. 

The group should have adequate capacity to assume responsibility for non-hospital 
physician services. It would be most preferable if the group practice had the ability either 
through its own physicians or through established referral mechanisms to provide all 
physician services covered in the performance standard. 

*■ Quality of care standards. It would be reasonable for HCFA to make efforts to assure 
that Medicare beneficiaries receive appropriate and timely care. Groups may be required 
to have an on-going quality assurance program to monitor specific clinical and non- 
clinical outcomes. Another desired feature would be a system to measure consumer 
satisfaction and complaints about services. 



50 



Other management systems could help to indicate quality of care. Groups might be 
required to have systems to monitor each physician's provision of care, backed up with 
standards for educating or removing physicians that do not perform according to 
established guidelines. It may be considered an advantage if the group practice had in 
place a system that provides for a uniform medical record for each Medicare patient seen. 
Rigorous credentialling procedures and appropriate board certification requirements for 
the group's physicians could enhance the group's ability to deliver quality care and 
enhance its standing as a GVPS candidate. 

► Patient management practices. There are other requirements HCFA may consider to 
assure that the group is capable of providing or monitoring care for all Medicare-covered 
services for the patients it sees. It is recommended that the group be able to demonstrate 
that its primary care providers are capable of serving the Medicare beneficiaries in 
multiple sites of service, such as offices, hospitals, nursing homes and the beneficiaries' 
homes. For example, the group's ability to institute and monitor the use of home health 
services, post-acute services in nursing homes, and durable medical equipment would be 
considered further evidence of its potential to succeed under GVPS. 

A system for following patients upon discharge from a hospital and for monitoring the 
health status of patients who are at a higher risk of hospitalization or judged to be 
medically unstable would be a desirable feature in terms of evaluating the group's ability 
to provide high-quality, efficient care. For example, groups might use case managers to 
oversee patients with specific illnesses over an extended period of time. Groups may 
want to demonstrate their ability to institute and implement treatment protocols, 
especially those relevant to the care of Medicare beneficiaries. 

HCFA could require a baseline indicator of patient management, such as a group's 
involvement with evaluation and management services for Medicare beneficiaries. This 
criterion may be set as a dollar threshold, which would also help to limit participation to 
larger groups. For example, groups may be required to have provided at least $100,000 
in Medicare-covered services for evaluation and management services. 

HCFA also could require a baseline level of involvement with Medicare patients seen, 
such as a minimum Patient Capture Ratio for total services, physician services, or 
evaluation and management services. For example, participating groups may be limited 
to those with at least a 20-percent share of services to patients seen. 

► Data and information systems. Part of the process of managing care, shaping practice 
patterns, and achieving economic success involves the timely collection and analysis of 
quality, utilization and financial data. Under GVPS there could be an exchange of data 
between a group and HCFA, in order to track utilization and report performance relative 
to target levels. Important types of information to process could include: 



51 



• Utilization rates for the group's Medicare patients in comparison to patients seen 
by non-GVPS providers, including visits, tests, procedures, and hospitalizations, 

• Hospitalization rates by diagnosis, length of stay and number of physician visits, 

• Utilization rates by physician and other providers, 

• Referral rates to other group physicians and to outside practices, 

• Revenues and expenses by type of service, 

• Mortality or hospitalization rates for selected diseases, 

• Utilization rates for selective preventive services (e.g., screening mammography 
and flu shots), and 

• Descriptive reports on all patient satisfaction surveys done and details on waiting 
times to be seen for different type of visits (symptomatic, non-symptomatic, 
urgent, etc.). 

The nature and extent of Medicare information that is shared with groups needs to be 
determined in light of any legal or ethical issues. Specific boundaries will be needed 
regarding how much detail is useful or permissible. For example, exchanging extremely 
specific information, such as utilization patterns for specific beneficiaries or specific 
competing providers, may not be desirable or even administratively practical. 

X. Provider Responses to GVPS 

Making available a GVPS option will lead physician groups to consider many decisions, 
including possible strategic responses. GVPS could give groups the incentive to manage their 
patients' care more efficiently. The strategies that physician groups use to achieve efficiencies 
are likely to be influenced by a multitude of factors, including characteristics of the organization 
and the market environments. The following section describes potential provider responses to 
GVPS, including: 

Deciding whether to elect GVPS, 

► Developing an action plan, and. 

► Managing performance 



52 



A. Deciding Whether to Elect GVPS 

The reward and penalty options implemented by HCFA under GVPS would influence a 
group's decision whether or not to participate. If there are potential rewards and relatively little 
risk involved, most or all groups may desire to participate. Under those circumstances, HCFA 
would probably need to specify clear eligibility criteria for participation in order to obtain a 
manageable number of candidates. However, options that could impose financial penalties on 
groups that failed to meet performance standards may cause a substantial proportion of groups to 
decide not to participate. If participation requires concrete actions on the part of participants, 
such as investing in managed care systems, then groups may have to weigh the prospects of 
success against the expected costs of those actions. 

1 . Meeting Eligibility Criteria 

There are two levels of eligibility criteria suggested for groups willing to participate in 
GVPS (see Section IX). One level relates to administration and general feasibihty, and second 
level relates to apparent ability to manage care. To meet one or the other set of criteria, some 
groups may need to combine their efforts. For example, individual physician practices could 
pool their experiences to meet criteria related to size or scope. Similarly, physician organizations 
such as local or specialty associations may sponsor an arrangement by which physician practices 
could combine their efforts. 

Even beyond minimum thresholds of size for adequate stability of measures, larger 
organizations may be in a better position to make capital investments in managed care structures 
and personnel. Small groups could be less able to absorb the costs of managerial and 
administrative support and the increased start-up capital expenditures for new cost-saving 
technologies. Furthermore, in order to manage patients effectively and efficiently, groups may 
find it necessary to be in direct control of the resources needed to provide a wide spectrum of 
services. In this way, they would benefit from economies of scope. For example, 
interdepartmental linkages would facilitate the efficient triage of patients to appropriate service 
facilities and services. It is expected that groups opting for GVPS would have an incentive to 
expand the scope of their services through formal or informal mechanisms in order to provide a 
wider range of services within their own system. 

Another important indicator of a group's market position is its Patient Capture Ratio 
(PCR), which is the percentage of services (measured in dollars) provided to these patients 
directly by the group itself For a given level of RPUPS, a higher capture ratio means the group 
is collecting a larger proportion of dollars, and suggests that the group is better able to control 
more of the patients' services. 



53 



2. Service Categories for RPUPS 

Groups may have the option of electing GVPS for MVPS services only, or for all 
Medicare services. Reimbursement amounts, potential savings, and therefore potential rewards 
are larger for all Medicare services. All-inclusive performance standards may be appropriate for 
large groups with integrated delivery systems. Performance standards based only on MVPS 
services may be preferred by some smaller or less integrated organizations. However, the 
definition of services included in the PGR could affect the attractiveness of GVPS depending on 
the scope of services. For example, if the physician group was part of an organization that 
owned a hospital, including hospital services in the PGR could create financial incentives to limit 
facility costs through lower admission rates. 

3. Reward and Penalty Structure 

Financial incentives provide a powerful rationale for groups to expedite the adoption and 
incorporation of methods for providing more efficient care. The more that groups can anticipate 
receiving direct financial rewards, the more they may be expected to invest in innovative 
strategies to be more cost-effective. In addition, provider response to the reward and/or penalty 
structure also might be influenced by the potential size of the penalties. Some groups may 
choose to assume the risk of a penalty if the potential for reward would be even greater. 

B. Action Plan 

Depending on the eligibility criteria, groups participating in GVPS already may be at or 
near the cutting edge in terms of having the infi-astructure and managerial expertise to plan for 
and deliver efficient care. However, under GVPS, they might re-consider their current situation 
and recognize potential opportunities for making further gains in cost-effectiveness and/or 
market share. This section discusses some of the actions a group may plan to undertake under 
GVPS. 

Using average utilization and patient capture ratio as two dimensions describing groups 
in the Medicare market, a group can be classified according to the following matrix: 





MEAN RPUPS DOLLAR VALUE 


PATIENT 

CAPTURE 

RATIO 




low 


high 


high 


A 


B 


low 


D 


C 



54 



The PCR reflects the group's degree of involvement or control over services and 
expenditiu-es. Average reimbursement rates reflect two factors: the relative health status or need 
of the patients, and the degree of efficiency in service delivery. A group's location in this matrix 
at the start of participation under GVPS may affect its priority objectives and its strategies. For 
example, a group with a low capture ratio could take steps to raise the percentage of Medicare 
reimbursement dollars going to the group itself An overall strategy may involve intra- 
organizational and inter-organizational measures in both managerial and clinical areas. At the 
clinical level, groups may embark on strategies to change their product mix in response to market 
demand, modify scheduling to accommodate the needs of its patients, and change staff 
specialization patterns. Groups may embark on more joint ventures and plaiming in which they 
will share both production facilities and support staff. 

Breaking down the PCR into different categories can fiimish additional useful 
information regarding utilization for groups. For example. Figure 1 1 shows the percentage of 
patients categorized in deciles of within-group utilization for a Group X, a Type B organization 
(high PCR, high RPUPS). Note that utilization refers to physician and supplier services only. 
This group's overall PCR was 69.1 percent. Over one-third of patients had all of their utilization 
within the group, and another 14.5 percent had more than 90 percent of their utilization within 
the group. Fully two-thirds of all beneficiaries ever seen at Group X had more than 70 percent of 
their utilization within the group. This information can be analyzed in conjunction with the 
v^thin-group reimbursements and RPUPS for each category of patient. For example, Figure 1 2 
shows the distribution of within-group reimbursements (darker, lower line) and RPUPS (upper 
line). Putting the two graphs together, we may infer that Group X provides a large amoimt of 
tertiary care. As described above, such a group's strategies include prioritizing alliances with 
primary care providers and vigorously pursuing internal measures to improve managerial and 
clinical efficiency. 

Group Y's PCR and RPUPS show patterns of a Type D organization, (Figures 13 and 
14). It's overall PCR is 21 .9 percent. The vast majority (86.3%) of beneficiaries receive 70 
percent or more of their care elsewhere. Viewing this information in conjunction with within- 
group reimbursements and RPUPS, we may infer that Group Y is essentially a provider of 
primary care for patients who in general appear to need less services than patients seen by Group 
X. As described above, such a group may undertake to improve informal or formal affiliations 
with providers of tertiary care, in order to better manage their patients' utilization that currently 
occurs outside their system. 

If the PCR is used in the reward structure, groups would have an incentive to manage 
more of the patient's care in order to benefit from the comparative efficiency of their system, and 
to increase their share of the reward. There could be similar incentives to see more patients, 
since this could increase net income and provide a larger base for generating savings and 
rewards. However, seeing more patients could, at least at first, tend to lower the average capture 
ratio because new patients might tend to ease into the practice over time. It is expected that 
groups will implement additional strategies to increase market share and reduce costs both within 
and outside their system. 



to 



Figure 11 

GROUP X 



Annual Within Group Utilization of Part B Services (in Dollars), 

Medicare Patients 



Percent of 
Within 
Group 
Utilization 



100- 
91 - 99 

81 - go- 
yi - 80 - 

61-70- 
51 - 60 - 
41-50- 
31 - 40 
21 - 30 
11-20 
1 - 10 



■;:;gi;i;;?';i:l>;::|V 


;,;|iip;v'il;-;iif.: 








13";;,;, '-i. ,;, v/.. 







677}- 



5A\ 



4.4J- 



3^8J 
3>J 



41 



5 



8.8j 



10 



14.5)- 



15 20 25 30 

Percent of Patients 



SOURCE: Analysis of 1992 National Claims History Data by Institute for Health Policy, Brandeis University 



35 40 

Pt. Capture 
Ratio: 69.1% 



Figure 12 



GROUP X 

m 

Mean Annual Group and Total Utilization of Part B Semces (in Dollars), 

Medicare Patients 



Mean 
Dollars 




Patient Categories by Percent of Within Group Utilization 

^ J ^ Pt. Capture 

Ratio: 69.1% 

SOURCE: Analysis of 1992 National Claims History Data by Institute for Health Policy , Brandcis University 



Figure 13 



GROUP Y 

Annual Within Group Utilization of Part B Services (in Dollars). 

Medicare Patients 



Percent of 
Within 
Group 
Utilization 



100 
91 - 99 H 
81 - 90 
71 - 80 
61 - 70 H 
51 - 60 
41 - 50 
31 - 40 - 
21 - so- 
il - 20 
1-10- 



1J> 



2.8 



5.9] 



379} 



6.1 



879} 



1 1 .6] 



16.1 



5 



10 15 20 25 

Percent of Patients 



SOURCE: Analysis of 1992 National Claims History Data by Institute for Health Policy, Brandeis University 



31.7 



30 35 

Pt. Capture 
Ratio: 21.9% 



Figure 14 



GROUP Y 



00 



Mean Annual Group and Total Utilization of Part B Services (in Dollars), 

Medicare Patients 



1600 



Mean 

Dollars .^o^i- 



1200 



1000 



BOO 



600 



400- 



200 




1 - 10 



1 1 - 20 21 -30 31 - 40 41 -50 51 - 60 61 - 70 71 - 80 81 - 90 91 - 99 



100 



Patient Categories by Percent of Within Group Utilization q^^^^^^ 

Ratio: 21.9% 

SOURCE: Analysis of 1992 National Claims History Data by Institute for Health Policy , Brandeis University 



59 



1 . Increasing Market Share 

No health care organization is impervious to the external environment in the current 
climate of cost-control. Under GVPS, groups are expected to vigorously study, explore, and 
analyze their current position in the marketplace, and formulate strategies according to their 
identified strengths, weaknesses, opportunities, and challenges. One goal could be to increase 
' market share among the Medicare population. 

The major strategy that groups use may be to market themselves in terms of cost, quality, 
and comprehensiveness of services. The strategy may be one of expansion in order to provide 
patients with a complete palette of services and to have more clout with purchasers of health 
care. They may engage actively in constructing strategic alliances with other providers of health 
care in order to provide a completely integrated continuum of care to patients. 

With such a strategy, integration is expected to take place horizontally as well as 
vertically, including across disciplines. Groups may concentrate on acquiring or allying 
themselves with primary care providers and creating a network of referring physicians. They 
may follow a strategy of mergers and acquisitions of previously free-standing facilities and 
service organizations, such as nursing homes, skilled nursing facilities, rehabilitation services, 
mental health and substance abuse treatment facilities, laboratories, visiting nurse organizations, 
ambulance services, pharmacies, and suppliers of durable medical equipment. 

The market strategies that groups prioritize or pursue more vigorously may be expected 
to be a function of both their RPUPS and PGR. According to our typology, a Type A group 
could resemble an HMO because of efficient practices and/or relatively healthy patient 
populations. They may be expected to focus on increasing their patient base, perhaps actively 
marketing themselves as high quality and efficient providers. Type B groups may also attempt to 
increase their patient base, perhaps by actively pursuing mergers with primary care providers and 
building a strong primary care referral base. 

Type C groups, with high RPUPS and low PGR, may be of two types: those with and 
those without a capacity to provide the necessary services to a patient population with significant 
and diverse health care needs. Those with the capacity could market their existing services to 
their patients, and make it attractive for these patients to remain within the system. On the other 
hand, those groups with patients leaving the system due to inadequate capacity or services will 
need to make strategic decisions regarding expanding versus strategic alliances. It might be most 
beneficial for such groups to affiliate with providers of tertiary care, who in turn would be 
looking for a dependable base of patient referrals. Finally, Type D groups have patient 
populations with low overall utilization rates as well as low participation within the groups. 
These groups have an opportunity to explore adding new patient populations to their group, and 
expanding the scope of their services accordingly. 



60 



2. Managing Cost and Utilization 

a. Within the Organization 

Many health care providers are competing on the basis of cost. Reductions in cost often 
are achieved within the organization responsible for providing health care. Physicians gain 
patients by discounting fees, organizations receive volume discounts from pharmaceutical firms, 
and so on, but beyond a point, cutting costs may affect quality and some say explicit rationing 
may be imposed. In order to counter such accusations, groups will need to employ tandem 
approaches to cutting costs and ensuring quality. Organizations will compete and differentiate 
based on quality, and not simply on cost. This involves such criteria as better access for patients, 
better process and outcomes, and increased customer satisfaction. 

An important component of reducing costs within the system is utilization review. 
Upgrading information systems in order to track concurrent rather than historical financial and 
clinical data is expected to become common practice. It is also expected that groups will 
upgrade their risk management and malpractice avoidance strategies. As with other health care 
providers, groups have several important reasons for ensuring the completeness and accessibility 
of their patient medical records. This may be more of a problem with group practices which 
have multiple sites where care may be delivered to the same patient. Groups therefore will have 
more incentives to create a single, composite electronic medical record made up of data drawn 
from all sites. Groups may build systems to construct lifetime clinical records, which in turn 
would lead to improved efficiency in managing patients seen at their facilities. In addition, 
groups may build an infrastructure in which patient referral and triage to specialists and other 
needed services is expedited and ensured. They could formalize on-call scheduling so that no 
needed specialty or service is ever unavailable, and no urgently required service goes unattended 
due to a lack of adequate and appropriate staffing. Groups are likely to develop and use group- 
specific standards of care, which can have legal standing and are therefore protective. 

It may be that health maintenance and preventive activities will play a larger role as 
organizations seek to control costs within their system. A problem is that beneficiaries are not 
committed to staying with the group.'" The payoff in terms of reduced costs for health 
maintenance and preventive activities is in the future, and therefore may not be recognized 
through lower RPUPS for that group. Patient loyalty is a factor that might counter this trend. By 
enhancing contact through health maintenance and preventive activities, groups may encourage 
patient loyalty. Patient outreach for education or follow-up may have the same beneficial effect 
of encouraging patient loyalty to the group. 



'"Because Medicare beneficiaries can disenroll from TEFRA HMOs at any time, they are 
not committed to staying with a provider in that context either. 



b. Outside the Organization 



61 



A significant challenge faced by physician groups will be to control costs outside their 
system. A certain segment of the patient population would be expected to travel, and in enrolled 
situations, organizations are able to control costs incurred outside their system by various 
mechanisms such as pre-admission certification, payment only for emergency care, and co- 
payments or deductibles for utilization other than at preferred sites.'' In non-enrolled situations 
such as those faced by physician groups under GVPS, these approaches may not be appropriate 
or feasible. However, we are exploring the possibility of allowing groups to embellish the 
services available to Medicare beneficiaries (e.g., through discounted pharmaceuticals) that may 
function like coverage limits to foster loyalty. Presumably, groups could finance the extended 
benefits through expected (or past) reward payments. 

At the individual patient level, groups will need to find ways of educating patients about 
treatment options outside the system in order that they may make informed decisions. For 
example, patients seeking procedures elsewhere could be given decision support. Patients who 
inform their providers within the system that they intend to travel could be supplied with travel 
advisory packets that include education about compliance with medications and sources of 
preferred care. A more global strategy would involve authorities outside the group, such as 
HCFA, that could profile providers in the area for appropriateness and efficiency of care, so that 
all providers in a market area could be held to a single standard of care. For example, 
gastrointestinal endoscopy and cardiac catheterization are two frequently mentioned procedures 
for which patients may leave the system in questionable cases. Providers of these services 
outside the system could be profiled and feedback supplied if utilization were higher than 
expected for that patient population. 

Again, where groups fit in the typology may influence the strategies they find most 
useful. Types A and B groups do not need to prioritize control outside their organization, since 
they already control a large proportion of their patients' utilization. Type C organizations may 
find themselves in the most vulnerable position of all, as they manage a group of patients who 
have high utilization that is incurred mostly outside the system. These groups may need to 
pursue most actively a strategy of strategic alliances with other providers of care for their 
patients, and include contractual arrangements for aligning incentives among themselves. 
Although the stakes are lower for Type D organizations because of their patients' low RPUPS, 
they too may be served best by encouraging affiliations and mergers with other area providers. 

Questions that all organizations face as they attempt to control costs outside their system 
will include the best affiliations that they may make. For example, if such arrangements do not 



''From the beneficiary perspective, an advantage of GVPS is that traveling and seasonal 
migration do not interfere with supplemental insurance benefits. HMOs may prohibit 
membership if a beneficiary is out of the service area for extended periods. 



62 



already exist, should they seek to affiliate with hospitals, primary care physician groups, 
specialty groups, and/or other types of providers? How formal should these arrangements be? 
What incentives should be shared, and what are the legal issues involved, especially surrounding 
anti-trust and restriction of trade? How can physician leaders safeguard against threats to quality 
that may occur in the pursuit of cost control? Finally, how can groups best manage the risk 
associated with having finite control over utilization outside their system? Groups may be 
served best by encouraging and facilitating utilization within their own system of preferred 
providers. 

3. Care Management Strategies 

Health care organizations have traditionally held a service concept that advocates practice 
management. Many of them have sophisticated systems for managing the resources necessary to 
run their practices and controlling practice costs. Needing to manage the entire patient in order 
to compete reflects relatively recent changes in the marketplace. Similar incentives would exist 
and be extended under GVPS. 

Some groups may utilize the gatekeeping concept in order to manage (but not necessarily 
restrict) patient access to specialty care. Assigning patients to a primary care provider could 
enhance patient loyalty as well. In certain cases, specialists can serve as gatekeepers to provide 
care that is more effective and efficient. Commonly cited conditions include end stage renal 
disease, diabetes, ulcerative colitis, and several cardiac and pulmonary conditions. Therefore, 
some groups may be expected to enhance specialist referral and consultation within their 
organizations, and develop efficient triaging systems. 

Under GVPS, groups will have an incentive to provide services in the most cost-effective 
way possible, and their recruitment and staffing mix is expected to reflect this. The appropriate 
use of physician substitutes has long been recognized as a legitimate method for reducing costs 
and improving patient satisfaction. GVPS groups also may expand the role of non-physician 
providers in reducing costs and fostering quality. For example, groups may use pharmacists to 
educate patients, fostering patient compliance (especially for chronic conditions such as asthma, 
degenerative joint disease, pain management, and diabetes), and thus reducing long-term costs 
associated with non-compliance. 

Groups may analyze the full range of health care needs of their patient population, and 
explore ways to fulfill these needs. Groups will need to consider "make or buy" decisions, i.e., 
whether to provide additional services internally, or rely on external providers. The volume of 
patients needing a particular service likely will affect which strategy an organization follows. 



63 



C. Managing Performance 

1 . Management Control 

Under GVPS, groups will have a greater incentive to manage the behaviors of individual 
providers and patients. Groups may enhance systems of bureaucratic control, such as protocols 
and guidelines, peer review, monitoring, training and education, and utilization review. Internal 
"market" controls also may be intensified, such as linking rewards and penalties to productivity 
and performance. Some groups may inculcate a form of control through group membership and 
its implications of living up to a certain professional and personal standard, adhering to a certain 
philosophy of care, and acting responsively to the mandates of organizational leadership. Some 
specific examples of how control may be exerted follow. 

a. Practice Guidelines 

Examples of clinical bureaucratic control systems are clinical guidelines, clinical 
pathways, standards of care, and practice protocols. Many already exist and are continually 
being refined as physicians and nurses gain knowledge fi^om recording their own experiences, 
observing and consulting with colleagues, and studying recent advances. However, there is often 
considerable variation among individual providers, and even from site to site. In order to achieve 
efficiencies, some organizations may develop, customize, and disseminate clinical guidelines on 
a continual basis. Groups may direct some clinical guidelines mainly at physicians. However, a 
solid organizational, cross-functional infrastructure is vital to the successful implementation of 
many types of interventions. The nursing profession has provided leadership in this area for 
many years. 

b. Feedback Mechanisms 

Many existing groups have feedback mechanisms so that physicians can evaluate their 
performance, both in absolute terms and relative to others. Currently, the nature of routine 
feedback to physicians varies widely. The trend is to provide utilization data, usually to the 
department head, who then shares it with individual physicians in the department. For example, 
physicians may receive information about the number of patients and case mix, number and 
types of procedures, average lengths of stay, and ancillary usage. Groups will need to provide 
accurate and up-to-date data to physicians on clinical, as well as financial, aspects of patient care. 



64 



c. Compensation Systems 

Physician compensation systems can be based on fees generated or a simple salary, with 
either mode linked to productivity, patient satisfaction, and other items such as teaching, 
research, and administrative responsibilities. Often the method of compensation reflects the 
culture and values of the organization, and changing the system can cause internal conflicts. 
Organizations that adjust compensation according to productivity have had difficulty when 
payers create opposing incentives. GVPS would align incentives under Medicare with incentives 
from other managed care payers. 

2. Evaluating Performance 

a. RPUPS 

The primary financial goal and performance measure under GVPS is meeting the target 
rate of increase in RPUPS. Unfortunately, precise measurement of performance will occur after 
the close of each performance year. However, GVPS is intended to evoke long-run changes in 
health care delivery patterns. Many of the management strategies required for success under 
GVPS are capital and systemic investments. Presumably, most existing organizations will have 
arrangements with many other payers that permit rapid assessment of short run as well as long 
run benefits. On the other hand, groups may find it helpftal to monitor their own practices, such 
as revenues per patient seen, to track performance periodically during the performance year. In 
addition, they can survey patients regarding utilization of other providers, perhaps in conjunction 
with patient satisfaction surveys. 

An important and unresolved issue for GVPS is the nature and extent of data sharing 
between HCFA and the groups. Even the minimum administrative requirements for HCFA are 
substantial, and adding frequent data reports could add a considerable extra burden. If groups 
are given information about market level trends in reimbursement rates, they can benefit from 
making comparisons to their own performance, especially if they also are able to analyze 
differential trends by type and place of service. 

b. Patient Capture Ratio 

Another critical success factor for groups could be the PCR. There are at least three 
possible reasons for a group to strive for higher PCRs over time. First, increasing the services 
provided in-house to the existing patient base translates into higher fee-for-service revenues. 
Second, groups generally will want to manage services internally, taking advantage of their 
integration and management controls. Third, the reward formula may include the PCR, meaning 
that higher PCRs lead to higher shares of overall Medicare savings. 

However, the PCR may convey some ambiguities. For example, if a group's Medicare 
patient population expands and new patients tend to have lower proportions of services within 



65 



the group, the PGR would decrease. Similarly, the PGR would fall initially if the group pursues 
a vigorous policy of cost and utilization control within its own system, with a lag in control of 
outside utilization. Furthermore, the PGR may not be an accurate reflection of the actual 
proportion of patient management by the group since it may fail to take informal alliances into 
account.'^ 

c. Services Provided 

Another way that groups might evaluate their performance is to monitor the types and 
volume of services they provide, as well as the sites at which they are provided. Groups might 
evaluate themselves on whether appropriate but less expensive sites are chosen for the delivery 
of services. An example would be to deliver sub-acute services in a skilled nursing facility 
versus an acute care hospital. When groups receive data on the proportion of services received 
within and outside their system, they can use that information to evaluate how successful their 
patient retention strategies have been, and in what clinical areas. 

XI. Simulations of Medicare Expenditures and Rewards to Groups Under G VPS 

In order to illustrate the flow of expenditures imder GVPS, this section simulates the 
overall financial impact of implementing GVPS on three parties: Medicare, GVPS groups, and 
other physician practices. We use a simulation model that incorporates baseline information 
about Medicare and the groups, along with a set of assumptions about the future evolution of 
Medicare with and without GVPS. We evaluate various scenarios by changing different 
assumptions in the simulation model. 

A. Analytic Framework 

1 . Defining Financial Impact 

For the purposes of the simulations, financial impact is defined as changes in total 
Medicare reimbursements for applicable services, plus any reward payments. Before proceeding, 
it is worth noting what this definition omits. For HCFA, reimbursements plus reward payments 
really do constitute Medicare's financial bottom line, although HCFA also is concerned with 
other impacts such as those affecting provider viability or beneficiary access. For providers, 
changes in reimbursement levels are a less exact measure of financial impact, since provider 
costs should be considered as well as revenues. For example, if reductions in volume cause a 
provider's revenues to decrease by $10 million, this does not mean that the provider's net income 



'*As mentioned earlier, an unresolved policy and legal issue is the nature of arrangements 
between a physician group and other providers that would permit the latter to be included in the 
group's PGR. 



66 



has fallen by $10 million. In reality, volume decreases would cause a decline in the provider's 
costs as well as revenues, although not necessarily in proportion. This consideration is more 
relevant over longer time horizons, where providers have more ability to implement changes in 
their cost structures. 

2. Time Horizon 

A time frame of five years is used for the simulations. For shorter periods, the effect of 
GVPS on the rest of the Medicare program would be less apparent, as there would be less time 
for interactions to come into play. 

3. Relevant Services 

The simulation model concerns the impact of GVPS on reimbursements for all Medicare 
services (Parts A and B). 

4. Defining a Base Case 

The approach taken is as follows: 

»■ A base case GVPS scenario is defined, comparing the distribution of reimbursements 
under GVPS to that which would be found without GVPS. 

► The sensitivity of the results to changing assumptions is examined. This involves starting 
firom the base case scenario and varying each major assumption in turn. For example, 
what if GVPS providers were paid a higher share of the savings achieved, holding other 
assumptions at base case values? 

B. Assumptions 

The simulation model includes three kinds of assumptions: 

Assumptions that must be made about the Medicare environment, regardless of GVPS, 

► Assumptions relative to the design features of GVPS, 

Assumptions concerning the likely impacts on provider behavior of a GVPS program. 

1 . Assumptions About the Medicare Environment 

Default growth rates. Even in the absence of GVPS, we assume that service volume 
growth will be more restrained for care delivered by the proposed GVPS groups, compared to 



67 



other providers. This is because groups that operate under GVPS are more likely to proceed with 
cost-reducing innovations because of other managed care incentive arrangements. 

We therefore assume that without GVPS, the volume and intensity of all Medicare 
services will increase by 10 percent for patients never seen by the GVPS groups. For patients 
seen at least once by these groups, we assume that volume and intensity growth will be 7.5 
percent for care delivered by the group itself, and 8 percent for care delivered to these same 
patients by other providers. The latter assumption reflects the idea that patients seen by the 
groups are being managed differently, and this difference also applies somewhat to the care the 
patients receive from other providers. 

Medicare enrollment and costs. We assume that Medicare has 30 million beneficiaries 
using services on a fee-for-service basis. For simplicity, we do not project any enrollment 
growth over the five-year period. Patients seen at GVPS groups are assumed to average $7,000 
in total Medicare reimbursements in the base year. For other beneficiaries, the average per- 
person reimbursement for all Medicare services is set at $5,000. These RPUPS values were 
chosen as representative of the differentials observed between our selected provider sample and 
their market areas. 

The difference in average reimbursements is large. However, this probably accurately 
reflects differences in case mix and service mix between the GVPS groups and other providers, 
which ultimately yield higher average costs. These differences in RPUPS levels are a source of 
opportunity for Medicare and the groups, in that large savings could result fi-om initiatives 
targeted at this high-cost population. 

Cost inflation factor. For the purposes of the simulation, we assume that the fees paid 
by Medicare for MVPS services are increased by a cost inflation factor each year. In the 
simulations, the same update for inflation of service costs is also applied to other Medicare 
services. We assume that inflation increases by 5 percent per year in all five years. 

Medicare Volume Performance Standards. We assimie that throughout the period 
considered. Medicare sets a target of 10 percent annual growth in volume and intensity. This 
assumption ensures that all providers meet the performance standard, since even the nongroup 
providers only increase their volume and intensity by 1 percent annually. Therefore, in the 
absence of GVPS, all providers would receive a full update (the cost inflation factor) to the fee 
schedule for MVPS services (i.e., no penalty). As a result, any changes in update factors 
observed in the simulations can be clearly attributed to the impact of GVPS. 

Expenditure Targets for Non-MVPS Services. The Medicare services currently 
included in the MVPS methodology represent a select set of Part B services (largely physician 
and other professional services). All other Part A and Part B services do not have target rates of 
growth with the size of future fee increases subject to penahies for excessive growth. In the 
simulations, however, we are examining scenarios where GVPS groups are evaluated for 



68 



managing the growth of average reimbursements per patient seen for all Part A and Part B 
services. Therefore, we assume in the simulations that HCFA has implemented volume 
performance standards for non-MVPS services analogous to the MVPS. 

Non-MVPS services then have a volume performance standard rate of growth of 1 
percent in the model. Since even the non-GVPS providers are growing at this rate for all 
services, all providers will operate within this target and be eligible for a full increase in payment 
rates according to the cost inflation factor. This update factor will be reduced only by the cost of 
GVPS reward payments. 

2. Assumptions about the GVPS Policy Design 

Computation of savings. Savings in reimbursements due to the GVPS policy are 
computed by comparing actual payments for patients seen by the groups to projected payments 
in the absence of GVPS. Projected payments are found by applying the specified volume grow^ 
target to the volume index for group providers in the base year. 

Annual rebasing of targets. An important design consideration is whether the volume 
performance targets are rebased annually to reflect the GVPS groups' year-by-year performance. 
This approach causes a "ratchet effect" with undesirable incentive properties which are discussed 
elsewhere in this report. Simply stated, a group that reduces volume not only loses 
reimbursements from forgone fees in the current year, but it also faces a tougher target in the 
following year. We present results both with and without armual rebasing of targets. 

Savings sharing rate. Rewards in the simulation model are computed in the following 
manner: 

Reward = Medicare Savings x Patient Capture Ratio x Sharing Rate 

In this formula the groups are not compensated for the fiill amount of projected savings. 
This approach is taken for two reasons. First, the savings amount will include reimbiu-sements 
forgone by other providers, not simply those forgone by the GVPS group. This follows from the 
fact that each group is evaluated according to how well it managed all care delivered to its 
patients, not just the care it provides directly. Applying the PCR to the savings amount corrects 
for this fact. 

Secondly, reimbursements forgone are an overstatement of net income forgone, as 
discussed earlier. The sharing rate therefore defines the percentage of reimbursements lost by the 
group through efficiency that will be paid by HCFA as a reward. The base case considered is 75 
percent, with HCFA retaining the other 25 percent portion. It is important to note, however, that 
HCFA also retains the entire amount beyond the group's attributed percentage (PCR). 



69 



Updates to fees for Medicare Part A and Part B services. In the absence of GVPS, we 
assume that HCFA updates fees for all services in accordance with the cost inflation factor, less 
any penalties for failing to meet volume performance standards. The maximum feasible update 
is therefore 5 percent per year, from our assumption about the cost inflation factor. 

In the scenarios, assume the fee update under GVPS is set as follows: 

Update factor = cost inflation growth rate 

Less (Adjustment to finance rewards to groups) 

We assume that the broad-based performance penalty is determined by evaluating whether the 
volume growth target was actually met by non-GVPS providers. The alternative would be to 
evaluate volume growth achieved by all providers. We do not recommend or model this option, 
because of its potentially undesirable effects. If the GVPS groups were included in the measure 
of national volume performance, then the more they restrained volume growth, the better the 
resulting measure of national performance would look. As a result, updates under GVPS would 
be higher for everybody than without GVPS, even if the non-GVPS providers failed to meet their 
collective targets. 

This would have two undesirable results. First, the incentive properties of GVPS for 
other (i.e., non-GVPS) providers would be dampened, as they could "free ride" on the efforts of 
the groups, and still get ample updates. Second, the cost to HCFA of paying higher updates to 
non-GVPS providers would reduce or even nullify the government's net savings from the GVPS 
initiative, casting doubt over the program's long term sustainability. 

In order to ensure budget-neutrality of the program, the proposed formula also includes 
an adjustment in order to finance rewards. The adjustment is designed to equate projected 
Medicare payments in the update year with GVPS (including reward payments) and without 
GVPS (using the fee update net of any performance penalty). In the simulations, actual volume 
growth is in fact equal to, or less than target volume growth. As a result, the only reductions to 
the update factor observed are due to the paying of rewards to GVPS groups. 

3. Assumptions About Providers Under a GVPS Program 

Patient Capture Ratio. For simulating the effect of GVPS, a key characteristic of a 
GVPS group is its capture ratio. For a given group, the PCR is its share of the annual utilization 
dollars for the patients it sees in the relevant year. A group that provides a little primary care 
(and nothing else) to many patients would likely have a low PCR when compared to a group that 
provides more primary care and/or costly tertiary care to its patients. 

Based on data that we have reviewed, we use a baseline PCR of 40 percent in the 
simulations, for all Medicare services. For some scenarios we consider a gradual increase in 
capture ratio over time, amounting to two percentage points a year. However, in the base case 



70 



scenario we assume no changes in the PCR in order to avoid over-complicating the interpretation 
of other effects. In order to enable the calculations to work out, the PCR does trend downward 
slightly over the five years. 

Reduction in service utilization and rates of growth. In the base case scenario, we 
assume that the groups will respond to GVPS and succeed in reducing the annual growth of 
service volume and intensity for their Medicare patients. This will include reductions in these 
patients' use of services from other providers as well. For care directly provided by the groups, 
utilization growth will be 1 percent lower than it would have been without GVPS (i.e., 6.5% 
annually). In addition, the groups will be managing the care their patients receive from other 
providers. We assume the success of this management depends on a group's PCR. Therefore, 
the GVPS groups with 40 percent capture will achieve only 40 percent of the reduction in 
utilization growth for externally-provided care that they achieve for the care they provide 
directly. In the base case, externally-provided care therefore grows at 0.4 percentage points 
below its non-GVPS rate of 8 percent (0.4 x 1% = 0.4%), so this volume grows at 7.6 percent 
annually. 

One of the most compelling features of the GVPS approach is the fact that GVPS groups 
have an incentive to reduce utilization in much the same manner that capitated HMO providers 
do, even though GVPS groups are still paid on a fee-for-service basis. Therefore, we anticipate 
that any group entering the GVPS program will move immediately to reduce, where clinically 
appropriate, the quantity of services utilized by the group's patients below existing levels. As a 
result, groups responding to GVPS do not only restrain the rate of growth in service utilization 
by their patients, but also achieve reductions in the absolute levels of service utilization. The 
base-year service utilization has some excess built-in due to the existing fee-for-service 
incentives. Groups deciding to participate in GVPS take steps to cut out this excess, since they 
will be rewarded for savings. In the simulations, therefore, we assume that GVPS groups cut the 
RPUPS for their patients 7 percent ($7,000 to $6,510) upon entering the program. 

For patients never seen by GVPS groups, volume and intensity for all Medicare services 
are assumed to continue growing at the default rate of 10 percent per year, with or without 
GVPS. This rate of growth is applied to the base year RPUPS level of $5,000. 

Increase in the proportion of Medicare beneficiaries seen by GVPS groups. The 

base case scenario assumes no growth in overall Medicare enrollment, with the percentage of 
beneficiaries seen by GVPS groups fixed at 10 percent. An alternative scenario allows for an 
expanded GVPS program with groups seeing an additional 15 percent of Medicare beneficiaries, 
so that 25 percent of all beneficiaries are receiving care fi-om GVPS providers for five years. 

For this alternative GVPS enrollment scenario, we assume that the additional patients 
have the same initial RPUPS as the groups' patient populations in the base case scenario. This 
implies that the GVPS groups are attracting higher-cost beneficiaries, and the average 
reimbursement outside the groups must be correspondingly lower. Therefore, in this alternative 



71 



scenario we assume that average reimbursements for patients never seen by GVPS groups (75% 
of the 30 million beneficiaries that use services) decline from an RPUPS value of $5,000 to 
$4,600. The issue of case mix is further discussed below. 

Change in patient case mix at groups. There is potential for some of the groups' 
initiatives to alter the mix of patients they see, leading to changes in their RPUPS levels. For 
example, a group might expand primary care offerings, thereby reducing an RPUPS which was 
previously high because of a focus on specialty care. To the extent that any risk adjustment 
methods failed to correct for such changes, there would then be a form of the biased selection 
problem familiar from HCFA's experience with HMOs. An important difference is that the 
groups would not be capitated, and therefore would face weaker incentives to favorably select. 
Like other fee-for-service providers, they face immediate loss of fee-for-service reimbursements 
if they avoid potentially costly patients, and this loss is unlikely to be outweighed by the promise 
of partial compensation in the future through the GVPS reward payments.'^ 

4. Summary: The Base Case Scenario 

As noted above, the simulation approach required a base GVPS scenario, which could 
then be varied on different dimensions. The base case GVPS scenario has been defined as 
follows: 

Medicare Environment 

1) 30 million beneficiaries use services each year. 

2) GVPS groups see 10 percent of Medicare patients. 

3) $156 billion in Medicare spending for all services. 

4) Cost inflation factor grows 5 percent annually for the 5 years. 

5) Volume performance standard for all services allows for 10 percent growth in service 
volume annually. 

GVPS Policy Parameters 

1) Medicare Savings under GVPS are the difference between actual payments for services 
to patients seen by GVPS groups and projected payments in the absence of GVPS. 



'^The Sharing Rate specified in the Reward formula would influence the strength of the 
• incentives. Simply adding relatively healthy patients could be a problem worth considering 
when monitoring results under GVPS. Appendix C discusses excluding from RPUPS 
calculations patients with low reimbursements to the group, effectively ignoring very low 
utilizers in the determination of Medicare savings and rewards. 



72 



2) Sharing Rate: GVPS groups get rewards equal to 75 percent of their share of the 
savings (i.e., Medicare Savings x Patient Capture Ratio x 0.75). 

3) Target rebasing: none over the five-year period 

4) The fee update factor is equal to the cost inflation growth rate, less the cost of GVPS 
reward payments. 

Provider Characteristics 

1) GVPS groups provide directly 40 percent (in dollars) of all Medicare-covered services 
that their patients receive (PCR=0.40). 

2) Volume of services provided to patients never seen by GVPS groups grows 10 percent 
annually, with and without GVPS. 

3) Prior to GVPS, the groups provide 7.5 percent more services per year to their patients 
while other providers provide 8 percent more services per year to these same patients. 

4) With GVPS, participating groups provide 6.5 percent more services per year to their 
patients while other providers provide 7.6 percent more services per year to these same 
patients. 

5) For patients seen by GVPS groups, base year RPUPS = $7,000. 

6) For patients never seen by GVPS groups, base year RPUPS = $5,000. 

7) Under GVPS, groups realize a 7 percent reduction in RPUPS. 
C. Findings 

1. Results 

Table 3 presents results of the simulation for the base case scenario, and contrasts them 
with projected resuhs in the absence of GVPS. Table 3 shows that under GVPS, the groups lose 
$1 .988 billion in reimbursements in year 5 (12.92% of what they would have received without 
GVPS). This loss is more than offset by a reward of $2,509 billion for their success in meeting 
targets. With the reward, the groups are 3.38 percent better off with GVPS than without. The 
non-GVPS providers also face lower reimbursements under GVPS, but the losses are spread over 
a much larger base, and therefore only account for 2.73 percent of their year 5 reimbursements 
without GVPS. Finally, Medicare saves 2.44 percent of total reimbursements for year 5 with 
GVPS ($7,709 billion), since higher payments to groups are more than offset by lower payments 
to non-GVPS providers. 



73 



Table 3 

Distribution of payments with/without GVPS 
Scenario: GVPS Base Case* 



Payments (SMedicare) in Year 5 


Scenario 


Difference (% change) with GVPS 




GVPS 


No GVPS 




Reimbursements to GVPS groups 
Reward payments to GVPS groups 


13,403 
2,509 


15,391 



-1,988 (-12.92) 
+2,509 


Total group income 


15,912 


15,391 


+521 (+3.38) 


Reimbursements to non-GVPS providers 


292,887 


301,116 


-8,230 (-2.73) 


Total payments by Medicare 


308,798 


316,507 


-7,709 (-2.44) 



Table 4 

Decomposition of GVPS 
Scenario: GVPS Base Case* 



Payments (SMedicare) in 
Year 5 


Change in Payments (Compared to No GVPS) 




To GVPS Groups 


To Non-GVPS Providers 


Total 


Reductions in Volume 


-1,700 


-2,595 


-4,295 


Reductions in Updates 


-288 


-5,635 


-5,923 


Reward Payments 


+2,508 





+2,509 


Total 


+521 


-8,230 


-7,709 



* GVPS groups see 10% of beneficiaries that use services in every year 



74 



Table 4 provides insight as to the source of the $7,709 billion savings for Medicare in the 
fifth year of GVPS, by decomposing this amount into price and volume changes. It may be seen 
that for the GVPS groups the larger share of the savings comes from reduced utilization, which 
in the model results from the GVPS groups' activities. Since the groups are managing all care 
for their patients, and non-GVPS providers account for 60 percent of this care, the non-GVPS 
providers also lose reimbursements as volume is reduced. In addition, both group and non- 
GVPS providers receive lower updates in order to finance the reward payments to the groups. It 
is worth noting that while all non-GVPS providers will experience the effect of GVPS through 
lower updates, the volume reduction effects will be experienced only by those who see patients 
managed by the GVPS groups. For the most part, these would be providers located in the 
groups' market areas. 

Table 5 addresses the effects of GVPS over time, by comparing volume and payments per 
beneficiary in year 5, v^th and without GVPS. Without GVPS, volume growth for Medicare 
over the five years is 60.5 percent, and payments per beneficiary increase by 102.9 percent. Note 
that even without GVPS, the groups' efficiency ensures that they experience slower volume 
growth than non-GVPS providers. With GVPS, the non-GVPS providers supply the same rate of 
volume growth as without GVPS (62.6%), and GVPS reduces their payment growth by reducing 
their updates. However, volume increases by only 30.5 percent 

Table 5 



Volume and Payments Change Over Time with/without GVPS 
Scenario: GVPS Base Case* 





Payments (SMedicare) 


Percent Change 


Volume per beneficiary 
(index) 


Base 
Year 


Years 


GVPS 


No 
GVPS 


GVPS 


No GVPS 


Beneficiaries seen by GVPS 
groups 


134.6 


175.7 


197.8 


30.5% 


47.0% 


Other beneficiaries 


96.2 


156.3 


156.3 


62.6% 


62.6% 


All beneficiaries 


100.0 


158.3 


160.5 


58.3% 


60.5% 


Payments per beneficiary 


Beneficiaries seen by GVPS 
groups 


$7,000 


$11,331 


$13,006 


61.9% 


85.8% 


Other beneficiaries 


$5,000 


$10,085 


$10,277 


101.7% 


105.5% 


All beneficiaries 


$5,200 


$10,210 


$10,550 


96.3% 


102.9% 



* GVPS groups see 10% of beneficiaries in every year 



75 



for the patients seen at groups, and payments for them increase by only 61.9 percent. As a result, 
payments for all beneficiaries increase by 96.3 percent over the five years with GVPS, which is 
lower than the growth of 102.9 percent that would have occurred without GVPS. 

2. Sensitivity to Policy Parameters 

Table 6 examines the sensitivity of our results to various changes in the policy 
parameters. It may be seen that increasing the sharing rule from 75 percent to 95 percent reduces 
slightly the total payments by Medicare in year 5. However, it increases the groups' gain fi-om 
implementation of GVPS, from 3.38 percent to 7.18 percent above their reimbursement total 
without GVPS. 

Of greater importance is the rebasing rule. The use of annual rebasing would make 
GVPS a money-loser for the groups, reducing their revenues 9.64 percent below the GVPS year 
5 amount of $15,912 billion in the base case. This reflects the ratchet effect of continually 
adjusting targets based on actual performance. However, the groups' loss in this case is not a 
gain for Medicare. Instead, the benefits accrue to non-GVPS providers, who receive higher 
updates (and therefore smaller revenue losses) than they would otherwise. This is because 
rebasing reduces measured savings, and therefore reduces the rewards to GVPS groups which 
would otherwise be financed through lower updates. 

3. Sensitivity to Other Assumptions 

Table 6 also addresses the sensitivity of results to varying the assumptions about program 
impacts. If the groups reduced volume growth to 4.5 percent instead of 6.5 percent, they would 
increase their income in year 5 by 1 .67 percent of the value without GVPS. The income gain is 
smaller than the 3.38 percent achieved in the base case scenario. This suggests that the 
additional rewards for curbing utilization more tightly are ultimately outweighed by the loss of 
fee-for-service reimbursements. With the 4.5 percent growth variation, Medicare would save 
2.93 percent of year 5 payments without GVPS, compared to 2.44 percent savings in the base 
case with 6.5 percent utilization growth. 

Alternatively, if the GVPS groups increased their PGR by 2 percent per year in addition 
to achieving the baseline utilization savings for Medicare, they would greatly increase their fee- 
for-service reimbursements. In this scenario variation, the groups' revenues in year 5 would be 
29.3 percent higher than without GVPS. For Medicare, this scenario resuhs in a 0.09 percentage 
point larger payment reduction than the GVPS base case because care is being transferred from 
other providers to GVPS groups, who are presumed to control utilization growth. 

Lastly, Table 7 presents the results of the enrollment scenario where GVPS groups see 25 
percent of the beneficiaries that use services, as compared to 10 percent of beneficiaries in the 
GVPS base case scenario. This alternative is labeled the GVPS 25 



I 



76 



Table 6 



Effect of Varying Program Impacts on Payments Under GVPS 
■ GVPS Groups see 10% of Beneficiaries that Use Services 





Varying Assumptions 


Payments (SMedicare) in 
Year 5 


Percent Change from No 
GVPS 


Scenario 


Groups' 
Utilization 
Growth 
Rate 


Annual 
Increase 
in 

Capture 


Groups' 
Share of 
Savings 


Annually 
Rebase 
Groups' 
Targets? 


To 
Groups 


To 
Others 


Total 


Groups 


Others 


Total 


No GVPS 










15,391 


301,116 


316,507 








GVPS Base Case 


6.5% 





75% 


No 


15,912 


292,887 


308,798 


+3.38 


-2.73 


-2.44 


Variants 




Vary Savings Share 


6.5% 





95% 


No 


16,497 


291,401 


307,898 


+7.18 


-3.23 


-2.72 


Vary Rebasing Rule 


6.5% 





75% 


Yes 


13,907 


296,665 


310,572 


-9.64 


-1.48 


-1.88 


Vary Utilization Growth 


4.5% 





75% 


No 


15,648 


291,570 


307,218 


+1.67 


-3.17 


-2.93 


Vary Capture Growth 


6.5% 


2% 


75% 


No 


19,900 


288,593 


308,494 


+29.30 


-4.16 


-2.53 



77 



Table 7 

Effect of Varying Program Impacts on Payments Under GVPS 
GVPS Groups see 25% of Beneficiaries that Use Services 





Varying Assumptions 


Payments (SMedicare) in 
Years 


Percent Change from No 
GVPS 


Scenario 


Groups' 
Utilization 
Growth 
Rate 


Annual 
Increase 
in 

Capture 


Groups' 
Share of 
Savings 


Annually 
Rebase 
Groups' 
Targets? 


To 
Groups 


To 
Others 


Total 


Groups 


Others 


Total 


No GVPS , 










38,478 


271,812 


310,289 








GVPS 25% Scenario 


6.5% 





75% 


No 


38,619 


252,647 


291,266 


+0.37 


-7.05 


-6.13 


Variants 




Vary Savings Share 


6.5% 





95% 


No 


39,713 


249,322 


289,036 


+3.21 


-8.27 


-6.85 


Vary Rebasing Rule 


6.5% 





75% 


Yes 


34,443 


261,139 


295,581 


-10.49 


-3.93 


-4.74 


Vary Utilization Growth 


4.5% 





75% 


No 


37,848 


249,572 


287,420 


-1.64 


-8.18 


-7.37 


Vary Capture Growth 


6.5% 


2% 


75% 


No 


48,082 


242,411 


290,494 


+24.96 


-10.82 


-6.38 



78 



percent scenario in the table. Expanding the proportion of beneficiaries seen by the more 
efficient GVPS providers yields even greater savings to Medicare. Medicare saves more than 
$19 billion in year 5 payments (6.13%), compared to spending without GVPS. In contrast to the 
GVPS base case, groups in the 25 percent scenario are only 0.37 percent better off in terms of 
fifth year revenues than in the absence of GVPS. Apparently, expansion in the size of reward 
payments has the effect of reducing fee updates to the extent that groups gain greater net 
revenues with 10 percent enrollment than 25 percent. 

Changing the policy parameters imder the GVPS 25 percent scenario has similar effects 
to those observed for the GVPS base case. Increasing the groups' share of savings from 75 
percent to 95 percent expands their net revenues gain from 0.37 percent to 3.21 percent ($1,235 
billion), and Medicare achieves a 0.72 percentage point increase in savings. Once again, the 
variation with annual rebasing of targets results in the groups receiving lower payments than 
without GVPS (a loss of 10.49%). Medicare's fifth year savings decline with armual rebasing 
from more than $ 1 9 billion to $ 1 4.7 billion. 

Table 7 also displays the result of assuming GVPS groups under the 25 percent scenario 
respond by limiting utilization growth to 4.5 percent rather than 6.5 percent. Under these 
circumstances, the groups end up with 1 .64 percent lower payments than in the absence of 
GVPS, while Medicare gains an additional 1.24 percent in savings. The last variation presented 
in Table 7 models the result of allowing groups to increase their PGR 2 percent per year over the 
five years. GVPS groups in this variation realize a 24.96 percent gain ($9.6 billion) in year 5 net 
revenues. Medicare payments in year 5 are approximately $19.8 billion below their projected 
level wdth no GVPS program. 

D. Discussion of Simulation Results 

The results in this section suggest that there are combinations of GVPS design and group 
activities that would generate gains for both Medicare and the groups. The most favorable 
scenarios are those in which the groups expand the proportion of Medicare patients they see, 
since this moves more patients into a slower-growing environment. As a result, both the groups 
and Medicare would benefit. Medicare is also affected by greater or lesser effectiveness of the 
groups' utilization controls, with Medicare gaining more savings as the groups control growth 
better. 

Policy parameters differ in their importance to the success of GVPS. Use of annual 
rebasing would reduce the groups' net income below the level in the absence of GVPS. 
Alterations in the sharing rule could also affect how well groups do under GVPS. Group 
participation could be deterred by overly restrictive policies in these two areas. 

The results in this section suggest that GVPS could potentially provide Medicare with 
billions of dollars in savings. The long-term impact of shifting beneficiaries from unmanaged, 
fast-growth settings to slow-growth, efficient providers could be of great significance to HCFA. 



79 



This policy could offer long-term gains beyond an immediate reduction in reimbursements. 
Quality, cost-effective providers will have an incentive under GVPS to provide services and 
manage the care of the most expensive Medicare beneficiaries. • 

XII. Conclusions and Recommendations 

Medicare has experienced large increases in spending during recent years, and most 
projections suggest these increases could continue. Policymakers have debated the merits of 
regulation, such as rate setting, versus market-oriented solutions, such as managed care. We 
believe that the regulatory framework established for MVPS is useful for achieving budgetary 
goals. However, we also believe it would be useful to supplement the current system with 
economic incentives that encourage the management of services. Many physicians and 
administrators who contributed to this study even commented that incentives under the 
traditional fee-for-service payment system — with or without MVPS — were out of step with their 
current efforts to manage care and improve efficiency. 

Managing services more effectively may elicit large savings from lower utilization rates, 
but bringing beneficiaries into managed care environments can be difficult. Medicare risk 
contracts with HMOs offer one opportunity. Capitation can create incentives for efficiency that 
encourage managed care. However, participation by an HMO is largely contingent on positive 
financial resuhs. In turn, enrollment by Medicare beneficiaries into an HMO is contingent on 
better benefits and/or lower premiums than competing Medicare supplemental policies. The 
current system pays HMOs 95 percent of estimated costs and therefore could save Medicare up 
to 5 percent for enrollees. Unfortunately, Medicare saves less than 5 percent, or even loses 
money, in cases where the average cost estimates are too high because of favorable selection. 

We hypothesize that Medicare could achieve greater savings from GVPS than from the 
capitation system: 

► First, the chances of Medicare losing money may be less under GVPS because the 

performance standards are based on the experience of the group. In contrast, capitation 
embodies "performance standards" that may have little correspondence to actual 
enrollees. Although there is always error associated with estimating expected costs, the 
experience of a group's own patients may be a more valid basis than the experience of 
other providers' patients. 

»■ Second, the financial benefits of managing care can be shared more evenly under GVPS. 
The formulas for sharing the savings can give ample incentives and rewards to groups, 
yet still allow Medicare to benefit substantially. Under capitation, any savings to 
Medicare are capped at 5 percent of mean reimbursement levels. Under GVPS, Medicare 
may keep the majority of savings for patients seen by most groups. 



80 



► Third, under GVPS groups have incentives to serve and manage expensive Medicare 
patients. Providers paid under fee-for-service are encouraged to seek and retain patients 
most in need of services. Capitated health plans have incentives to seek and retain 
relatively healthy members, not patients. 

We also see advantages of GVPS over state level MVPS and penalizing hospital medical 

staffs: 

► Our analysis suggests that increases in Medicare costs are more pronounced for Part A 
and other non-MVP S services. It would seem worthwhile to pursue comprehensive 
policies that embrace all services. Under GVPS, HCFA can follow reforms in the 
industry and encourage management of all services. Policies addressing MVPS services 
within states, or physician services within the hospital, are relatively narrow and not 
aimed at the major problem areas or most promising solutions. 

► Also worthwhile would be coherent polices that link appropriate incentives to the 
responsible decision-makers. We believe physician groups are the optimal focal points 
for comprehensive and coherent Medicare payment policies. Based on what we found, 
there are physician groups willing to accept the challenge. 

► In different ways, state level MVPS and hospital medical staff policy options are subsets 
of potentially more comprehensive GVPS policies. Setting regional or local performance 
standards is one necessary step in the process of establishing a GVPS option, which 
completes the process by giving incentives to providers to respond. Hospital medical 
staffs are potential candidates for GVPS, which could give them incentives to manage 
ambulatory and institutional services. 

HCFA has several parameters to consider for GVPS, which involve various tradeoffs. A 
significant decision is whether to stay with the scope of physician and supplier services only. 
Other decisions have to do with the level of complexity to build into the algorithms for setting 
standards and measuring performance. Also, decisions are needed about the appropriate balance 
between incentives to participate, and incentives for efficiency among those who participate. 

► HCFA has the administrative capacity and relevant data to implement GVPS for 
physician and supplier services, or for all Medicare services. Most of the administrative 
burden lies with the physician services because of their large numbers. Ironically, adding 
the other categories of services increases the administrative burden relatively little, but 
greatly increases the scope of the incentives and potential savings. We recommend 
basing GVPS on all services. For non-MVPS services, projections used in setting 
capitation rates for HMOs could also be used to set performance standards for groups. 

► There are a number of potential refinements and variations discussed in this report. (See 
Appendix C). Again, they would involve data and capacity that HCFA already has, but 



81 



would add to the number of steps. The value of methods to dampen stochastic effects 
must be considered in light of selected criteria for participation. Allowing medium sized 
groups to participate, for example, may add to the value of refinements. We recommend 
that health status adjusters be employed, but further consideration is needed about which 
categories to use. 

Groups should be given incentives for improving efficiency. These incentives could be in 
the form of rewards and/or penalties. Although penalties may strengthen incentives for 
efficiency, we believe that interest in participating would be greatly reduced by the 
prospect of losing money. Assuming that HCFA is willing to set cumulative performance 
standards, we recommend that concentrated penalties not be included. Failure to 
capitalize on an opportunity to manage care and earn rewards is itself a sufficient penalty, 
as is rising above a cumulative target and diminishing chances for future rewards. Giving 
positive incentives similar to capitation, and allowing HCFA to share in the savings, 
could reap significant benefits for Medicare and participating groups. 

Based on these considerations, we recommend that HCFA consider models with 
parameters such as the following: 

► Establish eligibility criteria, such as groups with primary care physicians and specialists 
serving about 8,000 or more beneficiaries per year, and a Patient Capture Ratio of at least 
20 percent for physician services. 

► For a policy beginning within the next year, say early in 1996, choose 1994 as the base 
year. 

► Measure utilization as RPUPS based on all Medicare services. 

>■ Establish performance standards using the counties where at least 5 percent of the group's 
patients reside. These are cumulative, meaning annual increases are applied to the 
previous target, not the most recent actual RPUPS. Adjust the performance standard each 
year for changes in case mix from the base year. 

► Pay successful groups a lump sum reward consisting of a refund for revenues lost from 
national fee penalties, plus 75 percent of the product of estimated Medicare Savings times 
the Patient Capture Ratio. 

► Penalize unsuccessful groups only through any applicable national fee penalties. 



82 



REFERENCES 

Eisenberg, JM, Doctors Decisions and the Cost of Medical Care, Health Administration Press 
1986. 

Heskett, JL, Managing in the Service Economy, Harvard Business School (1986). 

Holahan J, Zuckerman S, "The Future of Volume Performance Standards." Inquiry 30:235-248 
(Fall 1993). 

Miller RH, Luft HS, "Managed Care Plan Performance Since 1980." JAMA, May 18, 1994 - Vol 
271, No. 19, pp. 1512-1519. 

Omnibus Budget Reconciliation Act of 1 989 (OBRA). 

Physician Payment Review Commission. 1990 "Annual Report to Congress" Washington, DC. 

Physician Payment Review Commission. 1992 "Fee Update and Medicare Volume Performance 
Standards for 1992" Washington DC. 

Physician Payment Review Commission. 1993 "Annual Report to Congress" Washington, DC. 

Physician Payment Review Commission. 1994 "Aimual Report to Congress" Washington, DC. 

Porell FW, Tompkins CP, "Medicare Risk Contracting: Identifying Factors Associated with 
Market Exit." Inquiry 30:157-169 (Summer 1993). 

Tompkins CP, Wallack SS, Porell FW, van Reenan C, "Managing Medicare Physician Services 
through Volume Performance Standards" Report to HCFA: Institute for Health Policy, Brandeis 
University, Waltham, Massachusetts, July 1992. 

Wallack SS, Tompkins CP, Porell FW, van Reenan C, Volya Medicare, "An Analysis of Group- 
Specific Medicare Volume Performance Standards" Report to HCFA: Institute for Health Policy, 
Brandeis University, Waltham, Massachusetts, July 1991. 

Weiner, JP, Starfield, BH, Lieberman, RN, "Johns Hopkins Ambulatory Care Groups (ACGs); A 
Case-Mix System for UR, QA and Capitation Adjustment." HMO Practice, Vol. 6 No. 1 : (13- 
19); 1992. 

Welch WP, Miller ME, "Analysis of Hospital Medical Staff VPS" Report to HCFA: The Urban 
Institute, Washington DC, December 1994. 



A-1 



I, Appendix A: Consultations with Multispecialty Groups 
A. IHP Outreach Efforts with Group Practices 

Over the course of the GVPS research effort, Institute for Health Policy research staff 
have engaged in a series of contacts and exchanges of information with a number of group 
practices across the U.S. These contacts have included activities such as the following: 

IHP presentations at groups' coUoquia, 
Limited surveying of groups by IHP, 

A GVPS conference organized by IHP with group representatives attending, 
Conference calls with group leadership. 
Visits to group sites by IHP. 

The consultations with the groups have informed the GVPS research effort, providing insights on 
the role of Medicare in group practices, the organizational capabilities and dynamics of group 
practices, and potential obstacles to the implementation of GVPS. Because of explicit and 
implicit confidentiality considerations, specific identification in this appendix of groups and their 
representatives is limited. 

Creation of the National Advisory Committee on GVPS 

A half-dozen group practices (Lahey Clinic, Cleveland Clinic, Ochsner Clinic, Henry 
Ford Health System, Mayo Clinic, and Lovelace Clinic) meet on a regular basis to discuss 
common issues and challenges. Discussions between Lahey Clinic representatives and IHP staff 
about GVPS led to Lahey extending an invitation to IHP to make a GVPS presentation at a 
meeting of these six clinics. In addition, the initial GVPS report (Wallack et al., 1991) generated 
interest among members of the American Group Practice Association (AGP A). As a result, 
researchers at IHP made a presentation on GVPS at an AGPA national conference. Lastly, 
interest in GVPS among some additional group practices grew out of their existing relationships 
with IHP. Based on all these outreach activities, IHP identified a small set of providers that were 
interested in GVPS and willing to engage in consultations with IHP. These providers were then 
organized as the National Advisory Committee on Multispecialty Group r'ractices with the 
following members: 

Cleveland Clinic Foundation (Cleveland, OH) 

Fallon Clinic (Worcester, MA) 

Geisinger Clinic (Danville, PA) 

Henry Ford Health System (Detroit, MI) 

Lahey Clinic Foundation (Burlington, MA) 

Lovelace Clinic (Albuquerque, NM) 

Mayo Foundation (Rochester, MN; Scottsdale, AZ; Jacksonville, FL) 
Ochsner Clinic (New Orleans, LA) 



A-2 



Scott &. White Clinic (Temple, TX) 

Upper Hudson Primary Care Consortium (Warrensburg, NY) 

Dr. William Conway, Vice-President for Medical Affairs at Henry Ford Health System, serves as 
committee chairman. The committee has subsequently been renamed the National Advisory 
Committee on GVPS. 

Small-Scale Survey of the Advisory Committee 

Each member of the advisory committee received in 1993 a small set of questions 
prepared by IHP. The questions on these committee response sheets were designed to produce 
enhanced understanding of each group practice organization's size and structure, physician 
specialty composition, and the Medicare component of total caseload. IHP received responses 
for 8 of the committee's 12 sites. 

The majority of the responding sites operate as non-profit corporations. Almost all 
physicians in each group practice work full-time at the practice. The size of the physician staff in 
1993 ranged from 11 J to 995 PTEs, with a mean of 421. Over the 1989-91 period, these 
physician staffs grew at each of the responding sites (1-10% growth). All of the groups that 
completed the committee response sheet deliver a broad range of primary and tertiary services, 
and the physician staffs represent accordingly diverse arrays of specialties. 

In 1993, each responding group was either involved in contractual relationships with 
managed care organizations or negotiating a possible relationship. Four groups indicated that 
they were operating their own HMO. All of the groups had longstanding quality assurance 
programs with most now organized on the continuous quality improvement model. 

The delivery of medical services to Medicare beneficiaries represents an important 
component of each group's operations. For the year 1991, the Medicare proportion of total 
patient caseload ranged from 1 1 to 49 percent. Medicare reimbursements represented 18 to 49 
percent of total group revenues over the same period. 

1 993 Conference on GVPS 

IHP organized a conference on GVPS that was held in Boston on November 5, 1993. 
Representatives for 10 of the 12 sites attended, and these representatives held leadership 
positions in their organizations (chief financial officer, medical director, chief administrative 
officer, etc.). The HCFA project officer for GVPS also attended, along with another 
representative from HCFA's Office of Research and Demonstrations. This conference had two 
main purposes: (1) a presentation by IHP of a detailed report on the progress to date in the 
development of GVPS models, methods, and research; and (2) a discussion with group 
representatives in order to identify potential pitfalls and opportunities in the development and 
operationalization of GVPS. 



A-3 



The group representatives were excited by the potential of GVPS and the available 
research findings, such as the demonstrated ability to trace a group's aggregate Medicare claims 
via the I.R.S. employer identification number (EIN). They also expressed a keen interest in 
continuing to participate in consultations on the development of group-specific volume 
performance standards. A series of conference calls with leadership at each site did ensue, at . 
least partially as a result of this conference. In addition, researchers at IHP obtained a list of 
issues whose resolution would serve to guide the ongoing GVPS effort. 

Visits to Advisory Committee Member Sites 

During the course of 1994, IHP researchers made visits to 9 of the 12 group practices 
serving on the advisory committee. Our objectives in visiting the sites were two-fold: 

1. To detail progress on the project, present findings from data, and to educate senior staff 
regarding the proposed GVPS models. 

2. To gather information about the sites which would help us decide how feasible applying the 
various GVPS models would be at the sites. 

Some information was shared between Brandeis and the sites. The common goal was to develop 
a better knowledge base from which to operate. We realized as we conducted our site visits that 
a unidirectional gathering of information about these organizations on our part was necessary but 
not sufficient to properly support the building and implementation of GVPS models at 
multispecialty groups. Thus, an important component of our site visits became the educational 
process we undertook in explaining GVPS to the groups' senior clinical and administrative staff. 
Prior to every site visit, we distributed to each group a draft document that explained the 
background of GVPS, and discussed several models. This document served as the starting point 
for presentations and discussions at our site visits. In addition, we used this opportunity to 
present site-specific findings from our analysis of the Medicare database. As site visits were 
carried on during a time-frame of almost a year, our analyses became more detailed as the year 
progressed. Therefore, sites which we visited later in the year were apt to receive more findings. 

All groups received a core set of reports earlier in the project, before site visits were 
initiated. This first round of reports were site-specific, and analyzed a site's Medicare beneficiary 
population along two core dimensions: RPUPS and PGR (see Section X). We followed this up 
with conference calls with senior representatives at the groups to discuss the implications and 
value of these data. A general consensus emerged among the group practices that they would 
like to receive blinded copies of the specific data prepared for all other sites, and this anonymous 
set of data was sent in a second round of mailings. We received positive feedback from the 
groups regarding these data since they could then compare their own performance with other 
sites. For sites visited later in the year, we were able to provide them with data regarding their 
Medicare patients' utilization of services (both within their organization and external to it) by 
place of service and by specialty. 



A-4 



Groups recognized the value of knowing their PGR, since this measure illustrates the 
extent to which their Medicare patients are utilizing the services of providers outside the 
organization. Any group pursuing a proactive approach to managing more of their patients' care 
must move beyond simply making services available in-house. Indeed, one of the sites had 
realized the importance of patient satisfaction in increasing patient loyalty, and as a matter of 
routine performed customer surveys to determine how satisfied patients were with the group's 
physicians. The group's goal was to identify the factors responsible for producing an acceptable 
level of patient satisfaction, and then strategically address these systemic factors in order to 
promote patient loyalty (thereby increasing PGR). 

At another site, group leaders admitted that the need to manage PGR had not been 
perceived; historically, there had been enough FFS volume. Given the changing envirormient, 
they acknowledged that future strategies would include measures to shift more of their patients' 
utilization inside the organization. In operational terms, it would mean that more of the patient's 
needs would be identified in-house, and utilization of appropriate services within the 
organization itself would be encouraged, such as through enhanced access. One example 
mentioned was the potential creation of a telephone reminder system for patients with chronic 
diseases that would promote compliance with their treatment regimens (including visits and 
recommended prevention protocols). Such a service innovation would have the added benefit of 
improving quality of care. 

A quality improvement strategy adopted by a second group involved employing a team 
approach for the care of diabetics with each team led by an endocrinologist. Evaluation of this 
innovation determined that it had drastically reduced the need for hospitalization due to hypo- 
and hyper-glycemic reactions. In the context of GVPS, the group's leaders posited that the 
innovation in diabetic care had accomplished two further strategic goals: increased PGR for the 
group's diabetic patients covered by Medicare, and efficiently contained growth in RPUPS levels 
for these patients. 

Discussions with groups that displayed a high RPUPS level for their Medicare patients 
centered around ways that such organizations could implement strategies to more efficiently 
manage utilization within the organization. Many of the advisory committee group practices are 
already engaged in executing operating strategies to streamline the delivery of care, avoid 
duplication of services, implement clinical pathways and treatment protocols, and reduce 
variation by standardizing care. Again, very often quality improvement efforts provided the 
impetus for these strategies. 

At one of the groups, disease management strategies were developed and implemented to 
manage the entire care of patients with certain diseases. For example, more intensive 
management of patients with asthma extended the use of outpatient care services and actually 
reduced the incidence of emergency admissions due to asthma. Similarly, this site instituted a 
triage system for female patients with uncomplicated urinary tract infections that involved a 
detailed questionnaire administered by a nurse practitioner. This system reduced the length of 



A-5 



courses of treatment, and also lowered the use of specialist consultations, cultures, and expensive 
antibiotics. At another group, implementation of a decision tree for breast cancer treatment 
(mammography finding triggering a therapeutic cascade) resulted in more timely management in 
accordance with accepted and recommended clinical protocols. According to the surgeon who 
developed the tree in conjunction with a team of clinicians and administrators, the group now 
provides these patients with better outcomes and more efficient care. 

At the time of our site visits we uniformly found site representatives to be relatively well- 
informed about project development. Questions usually arose about the implications for 
individual sites and the utility cf participating in a GVPS program. At this point, we discussed 
with group leaders those elements of a site's environment, both internal and external, that might 
determine the GVPS model a group would prefer to participate under. We invariably found that 
through this process of mutual information exchange, site representatives grew to appreciate the 
opportunities and advantages of operating under some model of GVPS. 

B. Organizafional Characteristics of Advisory Committee Members 

We have outlined in this section some observations regarding the characteristics and 
capabilities of the group practices we have consulted and that are relevant to GVPS. 

Organizations and Environments 

Most of the groups are large and well-established, and characteristically display large 
patient volumes, organized clinical leadership, and explicit clinical standards. These factors 
provide the potential to effectively manage utilization and costs, and create an atmosphere in 
which GVPS may be successfully implemented. Members of our MVPS Advisory Committee 
are dispersed geographically across the nation, and also vary by setting (rural, urban, or semi- 
urban). Most are located in major metropolitan areas. In addition, most of the member 
groups have a lengthy history spanning a half cenmry or more, and are known for the niche 
they have carved out for themselves in health care services. Several of them have new 
facilities with state-of-the-art architecture and engineering, while some of them are in the 
process of upgirading or adding to existing strucmres. The majority of the groups are non- 
profit organizations. All of them have governing boards of trustees, and an executive branch 
predominantly made up of physician leaders. Several of the groups novv have TEFRA risk 
contracts, or are actively considering such arrangements. Similarly, every group is either 
involved in ongoing contracmal relationships with managed care organizations or negotiating a 
possible relationship. 

The external market environment varies substantially among advisory committee 
groups. Some are positioned in areas where there are several organizations of comparable size 
and capability, and potential patients have a wide choice of providers. Some are 
geographically located such that they are the only organization with their capacity, but they toe 
acknowledge that geographic isolation has become less of a barrier to patient drainage from 
their market area than before. All the groups are faced with an environment that is changing 



A-6 



rapidly and have had to accommodate trends towards managed care, mergers and strategic 
alliances. Some of them are in areas where purchasing units such as business coalitions are 
beginning to have an impact on health care service delivery systems in terms of accountability 
in outcomes and efficiency. 

Power and Control 

Essentially, the groups are physician led and dominated. Physicians function not only 
as clinicians, but also as administrators, supervisors, and strategic planners. The role of lay 
managers varies among advisory committee members. At some groups, we have observed 
very solid partnerships and tandem approaches between clmicians and non-clinician managers. 
Other sites have placed non-clinicians in strongly supportive managerial roles. Where strong 
alliances between physicians and administrators have been observed, we have typically seen a 
democratic allocation of power among the organization's physician leaders. Where the reverse 
is true, we have typically observed a relatively autocratic situation. None of the groups we 
consulted has assigned any control over clinical matters to non-physician managers, other than 
strategic planning for service mix. 

The advisory committee groups employ three control methods on their physicians, 
balanced in varying proportions. Bureaucratic control is exerted through the development and 
implementation of protocols and guidelines, peer review, monitoring, training/education, and 
utilization review. Market control is exerted through positive approaches such as rewards and 
incentives; we have not observed any groups, at this point, utilizing negative market 
approaches such as risk-sharing or penalties. Finally, we found "clan" control can be a very 
powerful method of control. At some groups, there is a very strong culture of membership in 
an elite group of physicians, and its implications of behaving up to a certain professional and 
personal standard, adhering to a certain philosophy of care, and acting responsively to the 
demands by physician leadership. At others, we found less evidence of a "team spirit." 

Service Concept 

Traditionally, most of the groups making up the GVPS advisory committee have been 
known for their specialty care. Several of them have national and international reputations as 
referral centers for tertiary and quaternary care. Until recently, building a primary care base 
has not been a priority for such groups. Even now, efforts to shore up the primary care 
aspects of these organizations are motivated in large part by the need to "feed" their 
specialties. Representatives at most of the groups expressed the belief that the changing health 
care environment may no longer be able to support the kind of "top-heavy" organizations they 
have successfully marketed themselves as in the past, in terms of specialty care. 

Similarly, most of the groups we consulted have had a service concept that advocates 
practice management. Thus, most of them have fairly sophisticated systems in place for 
managing the resources necessary to run their practices and controlling the costs of delivering 
specific services. The need to manage the complete patient and all required services— both to 



A-7 



compete in the marketplace as well as to be successful under Group Volume Performance 
Standards— is a relatively novel concept for these organizations. 

We have received mixed feedback from the groups regarding their attimdes towards the 
gatekeeper concept. Several groups maintain that gatekeeping, via a primary care physician 
that restricts access to specialty care, does not automatically produce more efficient care. 
They point out that speedier access to specialists may indeed promote more efficient care for 
more complex or unusual cases. Physicians at some groups argue that for specific categories 
of patients, specialist care is far more effective and efficient. The examples most commonly 
cited are chronic conditions such as diabetes, ulcerative colitis, end-stage renal disease, and 
congestive heart failure. Furthermore, many of the physicians interviewed stressed that the 
ready availability of specialist referral and consultation in their groups actually leads to more 
efficient and effective patient management since delays and missteps in diagnosis and treatment 
are minimized. 

Given that GVPS would apply only to the Medicare population served by the group 
practices, there is concern that patients' payer class could inappropriately influence their 
access to care. We believe that this is not likely to occur for several reasons. First of all, 
group physicians affirm that payer class usually does not become a factor in the clinical 
management of patients. Indeed, we interviewed a number of physicians who indicated that 
they could not determine a patient's insurance coverage from the medical record; only clinical 
information is presented. The advisory committee groups have a reputation for providing 
quality care that they would seek to preserve and promote. They have a strong incentive to 
avoid allegations of improperly restricting care. Finally, although GVPS applies only to the 
Medicare population, other patient populations served by the groups have increasingly come 
under some form of managed care umbrella. 

Operating Strategy 

The advisory committee groups vary considerably in the extent to which the services 
they provide are integrated. All of the groups, however, are interested in reducing 
fragmentation and offering a consolidated package of services to their patients. They recognize 
that establishing a solid, widespread primary care base is essential. Some groups have already 
achieved this base, having adopted this strategy early on, but most of them are in the process 
of doing so. Strategies to build primary care bases have ranged from acquiring existing 
primary care delivery systems, establishing new ones, hiring new primary care physicians, 
freezing the hiring of specialists, and training and deploying primary care physicians in-house. 
Most groups have also diversified into other fields such as home health and nursing homes. 

The preferred organizational strategy appears to be to set up a system where several 
satellite ambulatory care sites feed into a single tertiary /quaternary care site. Further, more 
and more services will be provided at the satellite sites. As an example, one group has a 
couple of mobile MRI units which rotate through their multiple satellite sites. Most of the 
groups, in keeping with national trends, have expanded service delivery in the ambulatory 



A-8 



sector. They have created formal or informal alliances with providers covering the entire 
spectrum of health care services, such as skilled nursing facilities, rehabilitation centers, and 
home health. In multispecialty groups such as those making up the GVPS advisory committee, 
patients tend to remain within the system for substantial portions of their care, but the potential 
remains for the groups to enhance patient outreach in order to manage more of their care 
within their own systems. 

Information Systems and Information Flow 

While all the groups acknowledge the importance of state-of-the-art information 
systems, there is wide variation in what they actually possess and currently utilize. As is 
typical for the industry, information systems are most developed in the inpatient arena, and 
least so in the ambulatory setting. Most of the groups have information system modules such 
as those for ordering and reporting the results of laboratory, pharmacy, and radiology tests. 
However, few have completely integrated these separate modules. With the exception of a 
pilot project at one group's satellite facility, none of the groups has a completely electronic 
patient medical record system, although most of them have computerized substantial amounts 
of clinical data. This appears to be motivated primarily by billing needs, and secondarily by a 
need to document for reasons of accreditation, quality assurance, utilization review, and risk 
management. There is also wide variation in the financial reporting systems groups have in 
place. Within a group, we have even observed different reporting systems being used for 
different areas— for example, in decisions regarding what constimtes a cost center versus a 
profit center. None of the sites has completely integrated clinical and financial systems. 

The namre of routine feedback to physicians also varies widely. The dominant mode is 
the provision of utilization data to the department head, who often then reviews it with the 
individual physicians in the department. Some of the groups do generate comparative statistics 
on crude measures of physician productivity— number of patients seen, case-mix, and length 
of stay, etc. It appears that very little, if any, outcome data is either generated or disseminated 
routinely. The rationale for physician feedback in most cases appears to be financial in namre. 

An associated issue is the flow of strategically-relevant information within a group 
practice organization. In particular, we were interested in learning how these organization 
implement internal policies in response to external pressures. We wished to learn how 
information was received, analyzed, assessed, and transmitted among members of the 
organization. The manner in which groups use objective and subjective information to 
formulate strategy and develop norms and policies indicates how the groups will function 
under a new policy such as GVPS. We found a fair amount of variation in the ways the 
groups process information. However, the advisory committee members are remarkably 
similar in the manner in which they differentiate between processing clinical versus non- 
clinical information. Thus, processing external information, such as practice guidelines 
developed by specialty societies, was seen as a purely clinical function. Most of the groups, 
report that clinical information is internalized and disseminated to clinical staff pursuant to 
staff consensus building and guideline customization. 



A-9 



In general, processing non-clinical information is a joint exercise between clinician- 
administrators and managers, at a seniority level that is consistent with the projected impact of 
the information. Thus, at most of the groups, policies regarding strategic alliances or 
mergers, for example, would be evaluated and decided upon at the organization's senior level. 
The information would then be disseminated within the ranks. 

Physician Recruitment. Reimbursem ent and Retention 

The groups are remarkably similar in their culture regarding the hierarchical rank, 
status, prerogatives and obligations of their physicians; although they can differ in how much 
physician autonomy is encouraged or permitted. Some of the groups are only now beginning 
to move away from a strongly individualistic tradition into a more administratively controlled 
tradition. 

Most of the groups appear to have no problems attracting top-quality physicians. It is 
unclear whether there are demogranhic differences, although one group finds that it attracts 
more women physicians because of generous flex-time arrangements. Most group practices 
have close affiliations with medical schools (one group has its own medical school) and 
graduate medical education programs, and these ties serve as one source for recruiting new 
physicians. 

The most common method of reimbursing physicians in the groups is by salary, while 
some mixed a salary with incentives. None of the groups have physician staffs reimbursed 
entirely on a fee-for-service basis. We have observed one instance where certain specialties are 
being reimbursed a percent of billings, and another where a partial capitation arrangement was 
being gradually phased in for primary care providers. There is variation among the groups in 
how starting salaries and raises are calculated. Some groups use published ranges of salaries 
as guidelines for various specialties. Others either reduce or eliminate the free market 
differential among specialties. In all groups, seniority is a factor in calculating compensation. 

There are varying levels of sophistication in using productivity as a factor in 
reimbursement". Productivity may be cited as a factor in calculating year-end raises or 
bonuses, -as are teaching, research, and administrative and collegial responsibilities. We have 
not observed any instances of a withhold on physicians' compensation. A minimal risk- 
sharing arrangement was observed in one instance, although some groups have begun to 
initiate risk-sharing at the department level. Groups vary in regard to whether salaries or the 
method of calculating them are made public. The authority to set salaries also varies by 
group, being vested alternatively in one single individual, in the heads of departments, or m an 
organization-wide salary committee. 

When questioned, most physicians working at these groups say that the major reason 
they chose to practice there, in preference to other settings, is because of their organization's 
reputation for providing excellence in care. Controllable lifestyle issues are also an important 
reason for choosing to practice in a group setting. Physicians in these multispecialty groups 



A-10 



enjoy greater predictability of working hours along with freedom from administrative and 
bureaucratic duties such as hiring and firing, billing, and collecting. They cite the benefits of 
having regular hours, consistent caseloads, and adequate coverage. In addition, many 
physicians refer to t^^e enhanced oppormnities for patient referral and consultation provided by 
a group practice setting, particularly when the financial considerations of the individual 
physician are minimized. Many physicians enjoy the ability to combine an active clinical 
practice with applied research and educational activities that would otherwise be infeasible. 
The groups largely have very modern facilities and equipment, which are attractive to 
physicians. Some physicians cited the ready availability of legal and ethical consultation as a 
benefit of group practice. Group physicians appreciate the fact that they do not need to make 
the investment necessary when initiating a private practice. 

Most physicians interviewed acknowledge that the benefits of group practice come at 
the price of ceding some degree of autonomy. The structure of group practice facilitates 
greater peer regulation and oversight as compared to solo practice. We found that although 
group physicians recognize both the costs and benefits of this diminished autonomy and choose 
their practice settings accordingly, there is an underlying tension that occasionally remains. 
One issue that frequently provokes physician discontent is that most physicians at our groups 
have little control over patient load and scheduling. Salary and fair reimbursement for effort 
is also a topic of discussion, and sometimes dissension, that may arise among physicians in 
group practices. We have occasionally observed a dissonance among physicians and 
administrators (both clinical and lay managers) regarding what constimtes appropriate work- 
load and adequate reimbursement for physicians. Although we expect that some level of 
physician discontent regarding compensation will always remain, group leaders believe they 
can and will move to alleviate the tensions and uncertainties associated with this issue. 

Use of Phvsician Substimtes 

There is wide variation among the groups in the extent to which they routinely use 
physician substitutes and extenders. Some groups use nurse practitioners and physician 
assistants extensively in almost all departments whenever possible, while others have only now 
begun to-initiate this process in a very limited way. Leaders at all groups, however, recognize 
the value" of utilizing non-physician providers where appropriate. We received the unpression 
that physician resistance to the use of substimtes was not a factor affecting then- use within our 
groups. However, we were not able to determine whether patient resistance was a factor. 

Research and Education 

There is variation among the groups with regard to their participation in research and 
education activities. For some, such efforts constimte major centers of excellence for their 
organization. Some organizations operate significant programs for undergraduate and graduate 
medical education, nursing education, and the training of other health care workers. The 
participation of other groups in such activities is limited to providing facilities for affiliated 
educational instimtions to send their smdents. Similarly, some organizations are 



A-11 



internationally and nationally recognized for their clinical and biomedical research, while 
others are engaged exclusively in the provision of direct health care services. 

C. Managing Health Services Organizations and GVPS 

Based upon the knowledge we gathered from the site visits and other consultations with 
the groups on the advisory committee, we have used the framework developed by Heskett 
(1986) to outline below some elements of strategic orientation groups will have to adopt to 
operate successfully within the dynamics of Group Volume Performance Standards. 

Service Vision 

In order to effectively manage patients' care under GVPS, Health Services Organizations 
(HSOs) will need to develop a new strategic service vision that retains several features of both 
traditional fee for service (FFS) practice and managed care arrangements (MC). For simplicity, 
in this discussion we use the term managed care narrowly to refer only to HSOs that deliver 
services to an enrolled, capitated population. Of course, in the industry managed care can refer 
to multiple modalities in terms of financing, risk, and population served. Under GVPS, HSOs 
face the following situation: 

HSOs deliver care to a non-enrolled population of Medicare beneficiaries on a FFS 
basis. These beneficiaries cost Medicare an annual amount of dollars known as 
Reimbursements Per Unique Patient Seen (RPUPS). The proportion of RPUPS accruing to the 
HSO is known as the Patient Capture Ratio (PGR). A high RPUPS results from high 
utilization of services by a patient population, and is an indication of both case-mix complexity 
and the degree of efficiency and appropriateness in delivering services to this population. 
Thus RPUPS need to be decomposed into its two components: (1) reimbursements for services 
that are needed and clinically appropriate, and (2) reimbursements for services that are 
clinically inappropriate or provided at excessive cost, and are therefore inefficient. 

A high PGR indicates that the HSO has a high level of control over the beneficiary's 
utilization of services. Under FFS, HSOs will want a high RPUPS, but PGR may not be a 
factor as "long as volume of patients is high (number of beneficiaries served). For example, as 
long as enough patients are receiving services so as to generate sufficient revenue for the 
group, capmring more of each individual patient's utilization may not be a necessary goal for 
the group. In contrast, HSOs under MG will want a low RPUPS (utilization of services) and 
high PGR is assured by enrollment. Furthermore, increasing the number of beneficiaries 
served under MG is not always sought by HSOs since each additional beneficiary represents an 
unknown risk in terms of utilization. Under the GVPS model, HSOs will want both a high 
RPUPS (due to patient need and not inefficiency) and a high PGR. They will seek to increase 
the number of beneficiaries served as well. Table 1 summarizes the strategic service visions 
for FFS, MG, and GVPS. 



A-12 



TABLE I: STRATEGIC SERVICE VISIONS 





RPUPS 


PGR 


NUMBER OF 
BENEFICIARIES 


Need 


Inefficiency 


FFS 


high 


high 


+/- 


high 


MC (HMO) 


low 


low 


high 


+/- 


GYPS 


high 


low 


high 


high 



More GYPS advisory committee groups fit the FFS model versus the HMO model. Of those 
that fit the FFS model, there are differing degrees of how well positioned they are to enact a 
GYPS service vision, since there is a need to have systems in place to ensure efficiency in 
managing utilization. 



Target Population 

Under FFS, it can be said that populations target providers. Studies show that patients 
of FFS providers tend to be sicker with more complex problems, and need and utilize more 
services. MC patients tend to disenroU and in fact are sometimes encouraged to do so by their 
HSO when their condition suggests a need for increased levels of utilization. FFS providers 
have financial incentives to provide services for sicker patients since they warrant both an 
increased volume and intensity of services. On the other hand, MC HSOs have incentives to 
target populations that have less potential for high utilization. Under GYPS, HSOs have 
incentives to target high utilizers since they still receive FFS reimbursements to treat patients. 
Table II summarizes these concepts. 



TABLE II: TARGET POPULATION BY RELATIVE HEALTH CARE NEEDS 





FFS 


MC (HMO) 


GYPS 


Utilization 


High 


Low 


High 



Currently, most of the groups we consulted are tertiary care centers and accordingly 
attract patients with multiple needs and high service utilization. Those with more integrated 
systems may also have a proportion of patients who are lower utilizers (for example, the 
groups with primary care satellites). One site with a long history of involvement in managed 
care surprisingly turned out to have patients with low utilization at the site itself, but high 
utilization elsewhere. Two other sites with managed care experience had intermediate utilizers. 
We anticipate that these profiles of the groups will change under GYPS. In particular, we 
expect to see these groups increase their capture ratios. 



A-13 



Service Concept 

In his analysis of service industries, Heskett (1986) described an organization's service 
concept as "the way an organization would like to have its service perceived by its 
customers." Any HSO operating under GVPS will want to project the following message to 
its Medicare patients: 

We are your source for high-quality health care. You are our patient. We 
know you, both your medical history and your family simation, and we care 
about you. We will make every effort not only to speed your recovery from 
illness, but also to keep you well in the first place. 

The HSO will need to work to build a bond between the organization and the patients it 
serves. The patients must believe that the HSO appreciates their condition and needs, and that 
the HSO is doing what's best for them. Establishing this trust in patients is crucial for HSOs 
to succeed under GVPS. After all, the patients are free to seek services from any provider in 
the United States. If a physician at an HSO tells a patient that a particular test or procedure is 
not necessary, there are no rules or payment restrictions in place (unlike HMOs) to prevent 
the patient from obtaining that service elsewhere. Therefore, effective control of utilization 
under GVPS requires patients to believe that their HSO will provide them with the services 
they need, at a high level of quality . Trust is also needed to convince the patient who feels 
fine that undergoing a certain test today (e.g, mammography) could avoid more serious 
complications down the road. The patient and HSO should be in a long-term relationship that 
promotes wellness, rather than only an episodic relationship that treats acute illnesses. 

Operating Strategy and Service Delivery System 

An HSO must have an operating strategy designed to achieve its service concept. 
Prioritizing of decisions regarding operations, financing, marketing, human resources, and 
control is essential. The HSO that is successful under GVPS has deployed its resources so that 
strong bonds have grown between the organization and the patients it serves. Accordingly, the 
HSO must strategically focus on attaining the following objectives in its operations: 

(1-) Patients must have timely access to services. Since the patients under GVPS 
have free choice of providers, the HSO must have adequate capacity and points of 
contact so that patients can have their health concerns dealt with promptly. Otherwise, 
they will go elsewhere for treatment. 

(2) Patients must have ready access to HSO facilities. As noted above, many HSOs are 
pursuing strategies whereby a central tertiary /quaternary care facility is linked to 
multiple satellite facilities. The satellite facilities themselves are an improved access 
route for patients, but the HSO must not overlook the difficulties faced by patients in 
traveling to the central site when medically necessary. Of course, many elderly and 
disabled patients have general transportation problems that should be kept in mind, 
including such matters as facility design and layout. 



A-14 



(3) The HSO must be devoted to providing the highest quahty care possible. Trust 
between patient and provider cannot survive and grow if there are patient doubts about 
the quality of care provided. 

(4) HSO physicians and other personnel must be both clinically competent and 
motivated to provide that "extra effort" in caring for patients. Excellent clinical 
judgment and skill is a necessary starting point, but those providing care at the HSO 
must remember that patient loyalty has to be earned. 

(5) In treating patients under GVPS, the HSO must pursue a strategy of aggressive 
patient outreach and disease management. The HSO cannot wait for patients to present 
themselves for treatment of acute episodes. Instead, the HSO should have enhanced 
contact and communication with patients in order to foster compliance with treatment 
regimens and head off the development of more serious problems. Patients should 
have regular points of contact with the HSO so that they do not become lost in the 
"system." 

Putting these strategic operating objectives (and others) into practice is obviously a challenging 
prospect for HSOs. However, we have found in our consultations with group practices that 
these groups are often already focused on such objectives, typically as a result of their ongoing 
managed care activities. 

Realization of operating strategies is dependent on the structure of the HSO's service 
delivery system. The service delivery system represents how the organization has decided to 
deploy its resources in order to achieve its objectives. Since clinical staff are so essential in 
health care, the HSO must have in place high-quality clinical staff, along with sufficient 
training resources to maintain their skills. Moreover, clinicians must have the right tools 
available to effectively manage the care of patients under GVPS. For example, information 
technology to support computerized medical records, test ordering, and prescribing of 
medications must be in place so that physicians can make the right treatment decisions for their 
patients in a cost-effective manner. Information links between an HSOs multiple sites are 
essential.- In addition, such technology can facilitate patient scheduling and 
monitoring/outreach activities. 

The HSOs facilities should not only have state-of-the-art medical technology, but also 
be designed to be user-friendly. This is particularly important for the elderly and disabled 
population: the necessary wheelchair ramps, elevators, signage, etc. must be in place. Ground 
transportation to both satellite and central facilities must be readily available and easy to use. 

Lastly, a successftil HSO under GVPS will be devoted to continuous quality 
improvement (CQI) and process re-engineering. Clinical and administrative staff in the HSO 
must view their participation in such activities as a vital part of their jobs. This includes 
efforts to devise, implement, and monitor practice guidelines and clinical protocols. CQI 



A-15 



activities put in place monitoring -systems that can identify system bottlenecks and capacity 
restraints. Re-engineering puts the HSO in the mode of always re-examining the 
organization's operations to unprove effectiveness and efficiency. Control of quality and cost 
is essential. 

Again, the groups that we have consulted are well aware of the strategic planning and 
implementation requirements for operating the kind of efficient and effective health care 
delivery system needed under GVPS, and several of them are industry leaders in such 
innovations. 



B-1 



I. Appendix B: Data 



A. Main Sites 



A Group Volume Performance Standards program must judge and reward participating 
Medicare providers on the basis of their efforts, and not because they are beneficiaries of a 
fortunate patient draw. Therefore, enrollment in the program must be predicated on certain 
requirements in terms of structure, organization, and size. The efficiencies under GVPS should 
be achieved through effectively managed care, not through favorable selection or concentration 
in a narrow range of specialtip"?. As a result, each participating provider group may be required 
to have a certain range in terms of physician specialties and vertical integration of its services. 
Presence of utilization review and quality assurance programs and other features associated with 
managed care may also be prerequisite. Finally, selected sites must be large enough to generate 
reliable and stable RPUPS. 

This research project has examined in detail twelve large physician group practices, all 
meeting the prerequisite organizational qualifications and all large enough to likely have the 
requisite stability in their RPUPS values. As members of the project's National Advisory 
Committee on GVPS, these sites consented to have their records reviewed and analyzed for this 
project, and also offered help in putting in perspective the results of our quantitative efforts. 
These sites, in alphabetical order, are as follows: 

• Cleveland Clinic Foundation (Cleveland, OH) 

• Fallon Clinic (Worcester, MA) 

• Geisinger Clinic (Danville, PA) 

• Henry Ford Health System (Detroit, MI) 

• Lahey Clinic Foundation (Burlington, MA) 

• Lovelace Clinic (Albuquerque, NM) 

• Mayo Foundation (Rochester, MN; Scottsdale, AZ; Jacksonville, FL) 

• Ochsner Clinic (New Orleans, LA) 
•. Scott & White Clinic (Temple, TX) 

•• Upper Hudson Primary Care Consortium (Warrensburg, NY) 

Data used in our analyses were obtained from the Medicare National Claims History file 
(NCH), the relational database maintained by the HCFA Data Center (HDC) which brings 
together the Medicare service records of all types which once had to be kept in separate files. 
Service types include the following sources of claims: 



• Hospital inpatient facilities; 

• Outpatient department facilities; 

• Skilled Nursing Facilities (SNFs); 

• Hospices; 

• Home health agencies; 



B-2 



• Ambulatoiy Surgical Centers (ASCs); 

• Physicians and other professionals; 

• Suppliers. 

Identification of patients at each study site is made possible through the use of the 
PROVTAX variable in the NCR's Physician/Supplier Part B file (representing the employer 
identification number or EIN), Although not required by HCFA for reimbursement, this field is 
completed for over 95 percent of submitted Medicare claims. By using PROVTAX as a marker, 
we can also determine the physicians (identified by their unique physician identifying numbers 
or UPINS) and physician specialties represented at each site. 

B. Other GVPS Candidates ' 

Although only our twelve main sites have expressed their interest in participating in a 
demonstration, it is our expectation that HCFA would be interested in similar analyses of other 
providers as well. To add further perspective to our analyses, we sought to identify other 
multispecialty providers within our mkin sites' catchment areas for whom volume performance 
could also be measured. It should be remembered that these other group practices represent 
additional individual examples, In no way should they be considered to form a comparison 
group or a set of controls. Nor should their behavior be interpreted as average or representative 
for the local area. 

Identification of "Other GVPS candidates was based on witliin-state data from the 
Physician/Supplier Part B file of the NHC database. Again, assignment of patient to provider 
was facilitated by the file's PROVTAX variable (representing the provider Employer 
Identification Number or BIN). Within-state files were employed in this step because we 
required a number of passes through the database. Within the NCH database, a pass through a 
state file can usually be accomplished overnight, while a national-based pass can have as much 
as a thirty-day turnaround time. In using only the state files, we made the assumption that large 
multispecialty group practices were identifiable from the volume and types of services they 
supplied to within-£.ate patients. Even a clinic with a large out-of-state clientele would reveal its 
multispecialty group practice structure when services provided to local patients were exammed. 

The first pass tlu'ough each state file identified the catchment area for the main sites in the 
state. A site's catchment area was defined as the union of all three-digit zip code areas from 
which the site enrolled at least 5% of its within-state patient load, Maps of each site's catchment 
areas are provided at the end of this appendix. In the second pass, all EINs with patients in the 
catchment areas were identified, and in a third pass all Part B claims were collected and matched 
to the proper BIN. By examining EI>f-level frequency distributions for number of patients, 
number of Unique Physician Identification Numbers (UPINS), and number of physician 
specialties, we developed descriptions of the size and composition of each EIN's practice. We 
then individually handpicked extra sites for our study. To be selected a site needed thousands of 
within-state patients, hundreds of UPINs, and a physician specialty list with either a good 



B-3 



representation of specific primaryxare physician specialties (e.g., family medicine, internal 
medicine, OBGYN), or had "multispecialty" itself listed as the most frequent specialty. Up to a 
total often GVPS candidates per catchment area were selected for study. 

C. Sample of Random Providers 

The providers described thus far are all large multispecialty group practices ostensibly 
possessing the structural and organizational advantages to bring certain efficiencies to their 
delivery of services. To facilitate comparisons to "average practices" within the catchment areas 
of our GVPS sites, we also selected a number of random providers. Sufficient physician groups 
were selected randomly from each catchment area to generate an associated total patient 
enrollment of at least three thousand users of services per area. Among the providers selected in 
this step were large single specialty groups, smaller multispecialty and single-specialty groups, 
and even some solo practitioners. Because a weighting scheme based on annual patient load 
was employed, practices in the first category were far more likely to be chosen than those in the 
last. We have reserved this provider sample for future work. 

D. Sample of Random Beneficiaries 

For each study year (1991-1993) one hundred percent of the patients of the providers in 
our study (i.e., main sites and other large providers) were used in the calculations of RPUPS for 
each site. To examine further the possible differences among main sites in terms of measures of 
volume and intensity relative to their market areas, we also constructed random samples of 
patients within each locality. In each main site's catchment area, we selected a random sample 
of beneficiaries of between fifteen and twenty-five thousand individuals. Each sample 
represented one to five percent of all beneficiaries in an area and were generated on the basis of 
the last two digits of beneficiaries' HICAN identifiers, under the assumption that such numbers 
are randomly assigned. 

E. RPUPS Calculation 

At the foundation of this project's quantitative analysis was the calculation of each study 
site's reimbursement per unique patient seen (RPUPS). For each site two such RPUPS measures 
had to be computed, one based on Part B physician and supplier services (but regardless of 
physician provider), and the other based on all Medicare-covered services (again, regardless of 
provider). The calculations and construction of datasets used for the rest of our quantitative 
analyses required two passes through the six files making up the fifth leg of the NHC database. 
In the first pass the above described EIN lists were used as finder files to identify the patients of 
each EIN. The original output of such runs were HICAN lists, but we recognized that Medicare 
patients can be assigned more than one HICAN and these output lists were sent through HDC's 
EBW Workbench to be unduplicated and reduced. This unduplicating process reduced patient 
counts at our sites by approximately 10 to 15 percent. Next, using the edited HICAN lists as 
finder files, a second pass through all NCH files was made in order to collect and aggregate 



B-4 



claim amounts for patients of each EIN. Only claims from providers in the Part B file were used 
in the computation of the RPUPS for physician and supplier services. Institutional claim 
amounts from five other files (representing Part A claims and Part B outpatient and facility 
claims) were added in to compute RPUPS for all Medicare-covered services. 

In generating RPUPS based on all Medicare-covered services, we computed separately 
the sum of the "pass-through" reimbursement (calculated as a state-based per diem times the 
length of the visit) and the reimbursement for organ transplant. These two amounts are not 
included as part of the recorded claim amount but are part of total Medicare services. Other 
reimbursements separately noted in the Part A files, such as for medical education and capital 
improvements, are already included in the claim amount field of the files. 

F. Claims Records 

For this study one hundred percent of service claims records of selected beneficiaries 
were accessed from the National Claims History file. Variables contained within these records 
include: 

For the patient : 

• Medicare Health Insurance Claim, which uniquely identifies the patient;' 

• Beneficiary residence ZIP Code; 

• Sex; 

• Birth date; 

• Reason for entitlement (aged, disabled, ESRD); 

For the provider : , 

• Provider type (physician, supplier, solo, group, etc.); 

• Provider ZIP Code; 

• Provider Tax Number (EIN); 
Performing provider UPIN; 

• Referring provider UPIN; 

• Provider specialty; 

For the service : 

• Procedure codes and modifiers; 

• Date of service; 



^This includes the Beneficiary Identification Code which identified the relationship between the individual 
patient and the primary beneficiary. 



Place of service; 

Type of service (medical care, surgery, consultation, etc.); 
Diagnosis; 

Medicare reimbursement amount 



B-6 




B-7 



THREE-DIGIT ZIP CODE MAP : 
MASSACHUSETTS : 

I 

i 




B-8 



THREE-DIGIT ZIP CODE MAP 
MICHIGAN 




B-10 



THREE-DIGIT ZIP CODE MAP 
NEW MEXICO 



865 

(AZ OFFICES) 




i 



B-U 




B-12 



THREE-DIGIT ZIP CODE MAP 
PENNSYLVANIA 




B-13 



THREE-DIGIT ZIP CODE MAP 



TEXAS 




B- 15 



THREE-DIGIT ZIP CODE MAP 
FLORIDA 



■ 315 (GA OFFICES) 




I. Appendix C: Alternative Model Specifications 

This appendix discusses potential modifications to the definition of RPUPS for a group 
operating under GVPS. HCFA could define RPUPS to exclude some beneficiaries seen by the 
group, exclude some reimbursements for beneficiaries counted in RPUPS, and/or account for 
patients' relative health status. 

A. Excluding Some Beneficiaries From RPUPS 

Starting with the population of beneficiaries seen by a provider during a year, it may be 
desirable to exclude some from calculations of RPUPS. Possible reasons for excluding some 
beneficiaries include: 

• Enhanced stability of RPUPS over time; 

• Reduced provider gaming potential; 

• Closer alignment of the performance standard with a provider's influence over patient 
care. 

A straightforward basis for excluding beneficiaries would be a minimum threshold for 
reimbursements to the physician group. With this approach, beneficiaries who have minimal 
contact with the provider (e.g., $120 or less in reimbursements) would be excluded altogether 
from the group's RPUPS measure. Inclusion of beneficiaries with low reimbursements to the 
group does not necessarily cause problems because the definition of RPUPS in the performance 
year would be identical to RPUPS in the base year. Accordingly, outside utilization occurring in 
the base year would be expected implicitly in the volume performance standard. 

Nevertheless, there could be other reasons for excluding the beneficiaries with the lowest 
reimbursements to the group. GVPS is intended to address efficiency in terms of patient 
management, which includes exercising influence over patients' use of other providers. In the 
case of beneficiaries with high reimbursement totals, almost all of which are to other providers, 
the concept of patient management is likely to be the most tenuous. Reasons for minimal contact 
with a single provider could be a flu shot, a single consultation, reading of lab results, etc' 
Including these beneficiaries in the value of RPUPS could obscure the success of efforts to 
manage patients more closely associated with the group's practice. 

Moreover, beneficiaries with very low reimbursements to the provider would constitute a 
"full" person in the denominator of the volume measure. Thus, yearly differences in the 
proportion of such "drop-in patients" might cause variations in the value of RPUPS for a group. 
Excluding the lowest cost beneficiaries also could reduce the advantage to a provider of 



' Grounds for exclusion could be based on type of service as well. For example, a beneficiary could be 
excluded if the only services utilized at that provider were related to lab services. 



c-2 



increasing their proportional representation in RPUPS. A systematic increase in the fraction of 
low cost patients would tend to decrease the value of RPUPS, possibly giving an incorr'^ct 
appearance of increasing efficiency.^ 

Finally, beneficiaries with low reimbursement totals across all providers need not be the 
focus of special policies for aggregate expenditure control. Omitting them from the measure of 
volume and intensity is therefore not considered a substantial problem. 

Figures C-1 and C-2 show the effects of removing beneficiaries with less than $120 for 
the 78 sample providers described in Appendix B. Figure C-1 shows the percentage decrease in 
patients from excluding beneficiaries with less than the minimum reimbursement threshold. The 
average number of beneficiaries decreases by 53 percent, to about 7,000. Because some of the 
excluded beneficiaries have low total reimbursements to all providers, and some have high totals, 
the effect on RPUPS could vary across providers. The average Patient Capture Ratio for each 
provider increases from 18 percent to 25 percent. Results for the sample providers are shown in 
Figure C-2. 

We illustrated this option with the $120 threshold. In a real system, the threshold could 
be a multiple of the conversion factor used that year for most physician visits. The threshold 
level might be chosen to achieve two objectives: i) a suitably low proportion of patients with 
high reimbursements outside and very low reimbursements inside the group; or ii) a suitably 
high average pafient capture ratio. The dollar threshold could be adjusted each year to reflect 
changes in the conversion factor under the MFS. 

B. Excluding Some Reimbursements From RPUPS 

The definition of RPUPS also can be varied by excluding some categories of 
reimbursements, even though the beneficiary continues to be counted. The main reason for 
excluding some reimbursements would be to enhance statistical reliability of RPUPS. 
Accordingly, we give most consideration to excluding reimbursements in excess of individual 
outlier thresholds. 

Although not tested by us empirically, there could be other circumstances in which some 
reimbursements are excluded from RPUPS. For example, it may be desirable to exclude 
reimbursements during the first year of Medicare coverage for very expensive and rare services. 
At the provider level, some services may be excluded, such as those emanating from a new 
department. For example, a provider that opens or acquires a new oncology center may suddenly 
treat a different category of patient. Services delivered in the new center may be omitted from 
RPUPS during the performance year until the utilization experience of the new center has been 
factored into the baseline RPUPS, and thereafter the target. 



^ This tendency also is addressed in the case-mix adjustments, discussed later. 



c-3 



Other reimbursements that could be excluded are for services that occur before the 
beneficiary's first visit in the year to the group. Arguably, a provider has little chance to manage 
or influence utilization that occurs before initial contact with the patient. A variation on this 
approach would include these reimbursements anyway, //the beneficiary was seen by the 
provider in the previous year. This option would be easier to administer since it would not 
require going back through the files to access service data. The applicable service bills for 
patients of GVPS sites could be tapped on the fly as they are received, because lists of relevant 
beneficiaries for such sites would always be up-to-date. 

Generally for a populafion, most reimbursements are paid on behalf of a small fraction of 
patients. The measure of average reimbursements per pafient (i.e., RPUPS) could be distorted by 
a variafion in the proportion of high cost cases during a particular year. Thus, it may be useful to 
exclude or to discount reimbursements for individual beneficiaries that are in excess of a 
specified threshold. This is somewhat like stop-loss reinsurance mechanisms that shield primary 
insurers from excessive risks due to uncontrollable factors, but its main purpose here would De to 
enhance stability in measurement. To a certain extent, health status risk adjusters could 
compensate for changes in the incidence of high cost patients. However, this aspect of the model 
is intended to dampen stochastic effects on the measure of total RPUPS for each group. 

Because fee-for-service payments would be made for all covered services, including 
when reimbursements are in excess of this threshold, there is a potential problem in that some 
actual reimbursements would be ignored or discounted when measuring the savings to Medicare 
under the GVPS. In other words, the level of true savings depends on reimbursements above the 
threshold as well as below the threshold. However, with reinsurance thresholds in place there is 
an equivalent problem that reimbursements to the site above the threshold in the base year would 
be ignored when setting the target. In any event, true savings should be measured using 
performance standards and observed levels that are not substantially altered by stochastic error. 
Thus, we explore the empirical question of whether model performance improves with high cost 
thresholds. 

There are two potential approaches to setting threshold values. First, an absolute dollar 
amount can be specified, such as $10,000. Reimbursements for any beneficiary in excess of 
$10,000 would be discounted or not counted when calculating the provider's average 
reimbursements per patient. Second, an amount can be specified for a given provider that 
depends on the distribufion of reimbursements for that provider's patients. Providers seeing 
patients with higher average reimbursements could have higher thresholds than other providers. 
Choosing threshold values would depend on the mean and variance of reimbursements per 
patient. We tested only constant absolute dollar values for all providers. 

In order to study the impact of a high cost threshold, we chose for this report the simpler 
approach of merely ignoring all reimbursements in excess of $1.0,000 for physician and supplier 
services, or $30,000 for all Medicare services. The $10,000 threshold affected approximately the 



c-4 



top 2% of beneficiaries chosen in the overall sample, and the $30,000 threshold affected about 4 
to 5% of beneficiaries. 

We chose to truncate the reimbursements over the threshold in order to illustrate the 
potential of this approach to reduce variability. However, reimbursements for any Medicare 
beneficiary in excess of the specified threshold could be discounted by some amount, such as 50 
percent. Each dollar of reimbursements (to all providers) for a beneficiary in that category in 
excess of the threshold would count as 50 cents in the RPUPS. This option may be desirable 
since cost consciousness, albeit at a reduced level, would apply to all services. Because the same 
rules apply for setting the target as measuring outcomes, the diminishing effect of the thresholds 
on mean values is counterbalanced. 

C. Health Status Risk Adjusters 

The measure "Reimbursements Per Unique Patient Seen" reflects the mix of patients seen 
by the group. Changes over time in case mix might be large enough to diminish the validity of 
comparisons between actual and target rates in the performance year. If so, it would be 
important to adjust either the target or observed RPUPS in order to standardize the health status 
distributions. 

The relative health status of each patient can be described in terms of risk classification 
systems. This study uses a two-tiered classification system, such that: 

• Beneficiaries were first categorized hierarchically by Reason for Entitlement to 

Medicare: ESRD, Disabled, Aged. A person with any claim during the year indicating 
ESRD entitlement was categorized as 'ESRD' for that year. Remaining beneficiaries 
with any claims indicating Disabled status were categorized as 'Disabled' for that year. 
All others were categorized as 'Aged' for that year. 

Aged and Disabled beneficiaries were categorized by ACG. This system classifies 
pafients into one of 5 1 categories based on all of the diagnoses observed for medical 
service claims during the year (Weiner, 1992).^ Table C-1 show^ a list of the patient 
categories defined by that system. 

A brief description of the ACG methods is offered here. The Johns Hopkins Ambulatory 
Care Group System (ACGs) is a case-mix system for categorizing patients based on their age, 
gender, and ICD-9-CM codes presented in claims over a given time period (typically one year). 
Initially, over 6,000 ICD-9-CM diagnostic codes are assigned to one of 34 clusters known as 
Ambulatory Diagnostic Groups (ADGs). A patient may be simultaneously placed into anywhere 



^ We basically took the current version of the ACG system "off the shelf." Later versions of ACGs, or 
entirely different classification systems, could be substituted. 



c-5 



from 1 to 34 ADGs. Those ADGs that are similar in regard to persistence or recurrence of the 
diagnoses contained in them are then collapsed into twelve categories known as Collapsed ADGs 
(CADGs). Next, Patients are assigned to clinically logical and mutually exclusive groupings 
known as Major Ambulatory Categories (MACs). Finally, statistical variance tet jiiques are 
used to split MACs into 5 1 mutually exclusive ACGs based on age, gender, and combination of 
ADGs. 

Depending on the data available and also the research goals of ACG patient 
classification, the range of ICD-9-CM diagnosis codes extracted from patients claims for 
inclusion in the ACG grouper algorithm can vary. We used all ICD-9-CM diagnosis codes (both 
inpatient and outpatient) recorded on Medicare claims to place beneficiaries in annual ACG 
categories.'' 

The mean reimbursements per patient within each cell were determined. The ratio of the 
mean dollars for each cell to the total RPUPS for the provider constitutes the relative average 
reimbursements for beneficiaries in the cell. For any base year, there will be a c° tain 
distribution of patients across the cells, and the weighted average of the mean dollar amounts for 
each cell is the value of RPUPS for the provider. 

Figures C-3 through C-5 show the proportion of patients and reimbursements per cell in 
1992 for one multispecialty physician group. The patient population seen by that provider in 
1992 was categorized hierarchically according to the reason for entitlement, and (except for 
ESRD patients) by ACG. Beneficiaries with ESRD entitlement accounted for 2.3 percent of 
patients (Figure C-3), 9.8 percent of reimbursements for physician and supplier services (Figure 
C-4), and 3.7 percent of reimbursements for all Medicare covered services (Figure C-5). 
Disabled beneficiaries accounted for 9.1 percent of patients (Figure C-3), 10 percent of physician 
and supplier reimbursements (Figure C-4), and 9.2 percent of all Medicare reimbursements 
(Figure C-5). Aged beneficiaries accounted for 88.7 percent of patients (Figure C-3), 80.1 



" The creators of the ACG System at Johns Hopkins University have provided the followmg guidelines for 
ACG-eligible diagnoses: 

Since ACGs were originally designed to predict the need for ambulatory health care resources, 
assignment of an ACG is usually accomplished using diagnosis codes from ambulatory claims/encounters. 
Historically, JHU has instructed most users of ACGs to select diagnosis codes from "face-to-face" ambulatory 
claims (e.g., claims where a physician performed an evaluation or management service). Since early in 1992 we 
have been experimenting with more inclusive diagnosis code selection strategies. 

We have assigned ACGs using all diagnosis codes that appear on a claim (including inpatient diagnoses) 
as well as diagnosis codes from all professional services rendered in noninstimtional (ambulatory) settings. While 
formal sensitivity tests have not been performed, we are confident that these approaches do not compromise the 
underlying assumptions of the ACG assignment process.... We have analyzed the explanatory power of these 
approaches and found them to be comparable to ACG assignment using only "face-to-face" ambulatory diagnoses. 
(Johns Hopkins ACG Case Mix System Applications Manual, 2nd Ed., July 1993, Johns Hopkins University, p. 
11-12). 



c-6 



percent of physician and supplier reimbursements (Figure C-4), and 87.1 percent of all Medicare 
reimbursements (Figure C-5). Among both the disabled and aged beneficiaries, there were 
concentrations of patients and dollars in relatively few ACGs. For both the disabled and aged 
beneficiaries, the laigest concentrations were in four ACGs (41, 44, 49 and 50), which represent 
situations involving multiple disorders (see Table C-1). Among the aged patients. Figure C-5 
shows that these four ACG categories accounted for large proportions of total reimbursements: 
10.5 percent (ACG 41), 16.1 percent (ACG 44), 32.6 percent (ACG 49), and 20.3 percent (ACG 
50). 

In a subsequent year, the distributions of patients and reimbursements across the cells 
could change. For example, there may be proportionally more ESRD patients. However, the 
performance standards are based on the group's experience and a specified rate of increase. 
Implicit in the performance standard, therefore, is the distribution of patients in the base year. 
Since ESRD patients are likely to be relatively expensive, we may need to adjust for changes in 
case mix. Either the observed RPUPS could be standardized to the distribution of patient in the 
base year, or vice v rsa. 

Table C-2 shows the effects of reweighting the case mix for the multispecialty group 
shown in the previous figures, using distributions of patients and mean reimbursements per cell 
from different years. The table contains three sections, reflecting observed mean dollar amounts 
per cell taken from 1991, 1992 and 1993, respectively. Within each of the table's three sections, 
the mean reimbursements per case-mix cell for Year a (1991 in the first section) are multiplied 
by the case mix actually observed in Years e and 6 (1992 and 1993). This yields estimates for 
RPUPS, and RPUPSq based on applying the 1991 (Year a) per-cell reimbursement rates to the 
case-mix cell distributions of patients occurring in 1992 and 1993 (Years e and 6). Thus, the 
actual RPUPS for physician/supplier services in 1991 was $1,810, while the estimated RPUPS 
for 1992 and 1993 were $1,873 and $1,891, respectively. Similarly, in the bottom section of 
Table C-2, the actual RPUPS for physician/supplier services observed in 1993 was $1,676 with 
estimates of $1,600 and $1,659 for 1991 and 1992, respectively. 

Across each row of Table C-2, the values of RPUPS for this provider ascend over time, 
suggesting that the proportion of sicker patients got larger over time. In other words, the 
distribution of patients tended to change in the direction of higher concentrations in cells with 
larger mean dollar amounts. Between 1991 and 1992, this change affected the RPUPS for 
physician and supplier services by about 3.7 percent (using 1992 mean dollar values per cell). 
For 1993, the cumulative change was about 4.6 percent. Comparing the results using 1991, 1992 
and 1993 mean dollars per cell reveals considerable agreement in the magnitude of case-mix 
changes over these years. This suggests that the relative mean reimbursement amounts per cell 
are fairly stable for the provider across the years. 

Table C-2 also reveals that changes in case mix had larger proportional effects on RPUPS 
values for all Medicare covered services. Using once again 1992 mean dollar values per cell, the 
estimated RPUPS changed by about 4.2 percent from 1991 to 1992. Between 1991 and 1993, the 



c-7 



cumulative change was approximately 5.4 percent. Comparing these results to those using 1991 
or 1993 mean dollar values per cell reveals again general agreement in the magnitude of case- 
mix changes over the years, with the trend over time toward a higher-severity case mix. 

Table C-3 shows a similar analysis of changes in case mix, but uses the modified 
definition of RPUPS in which beneficiaries with less than $120 in reimbursements to the 
provider are excluded. As before, the results are similar using 1991, 1992 or 1993 mean 
reimbursement values to define the relative costliness of patient categories. Imposing the 
minimum threshold seems to reduce the magnitude of the change in case mix between 1991 and 
1992, from 3.7 percent (Table C-2) to 2.3 percent (using 1992 mean dollar values per case-mix 
cell for physician and supplier services). However, this desirable effect does not occur for 
cumulative changes between 1991 and 1993. The observed changes increase, from 4.6 percent 
without the minimum threshold (Table C-2) to 5.5 percent. It appears that excluding this part of 
the distribufion does not necessarily improve the statistical reliability of RPUPS for physician 
and supplier services. 

Table C-4 carries out the same analysis using the high cost outlier thresholds, which 
truncate reimbursements for individuals in excess of $10,000 for physician and supplier services 
and $30,000 for all Medicare covered services. For physician/supplier services, there seems to 
be modest but consistent improvement using the threshold. Using 1992 mean dollar values per 
cell as in Table C-2, the magnitude of changes in case mix decreases by roughly one-half a 
percentage point for both the 1991-1992 comparisons (3.7 to 3.2) and the 1991-1993 
comparisons (4.6 to 4.2). This suggests that the influence of outliers can lead to differences in 
RPUPS across years that are not recognized by the risk adjustment system. 

D. Implications of Altemafive Definitions 

Figures C-6 through C-9 show results of provider versus market level increases in which 
beneficiaries with less than $ 1 20 in reimbursements to that provider were excluded from RPUPS 
measures. The random beneficiary samples were used to estimate market level changes, 
however, there is no specific provider referenced. In general, the data points tend to scatter more 
and the average absolute percentage deviations tend to increase, compared to results shown in 
Figures 5 through 8. For the selected physician groups as well, in Figure C-8 several points 
move quite a distance away from the origin — for instance. Groups K, M and N. For the 
following year, in Figure C-9, there also are big changes — notably for Groups K and L. Given 
the generally worse performance of the models that exclude some beneficiaries, these big 
changes may be considered suspect until any further vindicating evidence is obtained. 

Figures C-10 and C-1 1 show the 1991-1993 cumulative results for models that exclude 
beneficiaries with less than $120 in Medicare reimbursements to the provider. As before, the 
points tend to scatter more than without the minimum threshold, and the average deviations are 
greater. For physician and supplier services, the average absolute percentage deviation increases 
from 8.1 percent to 9.6 percent; for all Medicare services, the increase is from 9.5 percent to 1 1.9 



c-8 



percent. Most of the selected providers remain in the third quadrant in Figure C-1 1 . Groups L, 
N and O appear to perform worse with the threshold, while Groups K and M appear to perform 
better. Again, at this point it is unclear whether these significant changes are desirable given the 
poorer statistical performance of the model with minimum reimbursement thresholds. 

E. Adjustments for Changes in Case Mix 

The results presented so far have not taken into account possible changes in the case mix 
of Medicare patients seen by a provider. We anticipate that the mix of patients seen by a large, 
mature provider organization will exhibit less difference over time than exists between providers 
during any given year. Consequently, the pressure put on a risk adjustment system is likely to be 
less in this context than, for example, accounting for differences between local areas or enrolled 
populations. 

We investigate the effects of changes in case mix in two ways. First, we categorize 
Medicare patients into risk cells, reflecting different relative expected reimbursement rates. 
Second, we look at the influence of high cost outlier patients on the stability and validity of 
RPUPS. 

Each year we can analyze changes that take place in the distribution of patients across 
these cells. The proportions of Medicare patients seen may tend to shift toward (or away from) 
cells with higher relative costs, suggesting sicker (or healthier) case mixes. The observed 
RPUPS can be standardized to reflect the original (i.e., the base year) distribution of patients, in 
order to remove the influence of changes in case mix. An equivalent approach would be to 
standardize the target level to reflect the distribution of actual patients seen in the performance 
year. We might prefer the latter policy because savings estimates in the performance year 
depend on who the provider saw this year and not some previous year. 

To classify Medicare beneficiaries according to relative costliness, we chose two types of 
categories: Reason for Medicare enfitlement, and Ambulatory Care Groups (ACGs). In order to 
make case mix adjustments for a particular group over time, the following steps were taken: 

The patient populations in the base and performance years were 
partitioned separately into 85 risk cell categories. First, beneficiaries were 
separated according to reason for entitlement: ESRD, Disabled, and Aged. 
The Disabled were placed into 40 ACG categories, while the Aged were 
place into 35 ACG categories.^ 



5 There are fewer categories available than 103 (1 + 51 + 51) because some ACG categories include age 
criteria. For example, an Aged beneficiary cannot be placed in ACG 2 which only applies to patients ages 2-5 
years. 



c-9 



• For the hypothetical performance year, we determined the percentage 
distribution of patients across cells and the mean reimbursements for 
patients in each cell. These are the components of the actual value of 
RPUPS in the performance year. That is, the sum of the products for each 
cell in which the percentage of patients is multiplied by the mean 
reimbursement amount. 

• The observed patient distribution in the performance year was applied to 
the base year, using reimbursement dollars per patient in the base year. In 
other words, the mean reimbursements per cell in the base year were 
reweighted using the patient distribution in the performance year. This • 
results in a new, adjusted value of RPUPS for the base year that estimates 
what a hypothetical patient population for the group (i.e., that mirrors the 
performance year) would have experienced in Medicare reimbursements. 

To the adjusted value of RPUPS for the base year we applied a market 
level rate of increase. (The market rate of increase also was adjusted for 
changes in case mix, using the same two-tiered classification system). 
This is an adjusted performance standard that now reflects the same health 
status distribution that actually occurred in the performance year, and that 
only adjusts for changes in patients' health status relative to average 
changes for all local providers. 

Figures C-12 through C-17 show resuhs for adjusting for health status the performance of 
groups for the years 1991-1992, 1992-1993, and 1991-1993. Results are shown for all services 
and for physician and supplier services only. These figures depict deviations from the targets 
that are based on average rates of growth for the respective markets.^ For all services, the figures 
show three measures of performance for each group: 

Unadjusted RPUPS (Unadjusted: black bar); 

RPUPS adjusted for the 86 patient health status categories as described earlier. (HS 
Adjusted: gray bar); 

RPUPS adjusted for health status and also truncated at a maximum reimbursement level 
of $30,000 per patient (HS / Outlier: white bar). 

For physician and supplier services, the figures show only the first two measures of performance 
for each group, i.e., without the high cost outlier adjustment. The 10 groups are ordered from left 



'The MFS was implemented during the course of this 1991-1993 period. Local procedure codes were phased out in 
favor of CPT-4 codes. In addition, specialty-based differentials in physician payment were eliminated after 1991. The impact 
of these changes on any provider's rate of growth relative to the market is uncertain at this time. 



C-10 



to right (i.e., from 1 to 10) based on the unadjusted value of their 1992 RPUPS for all services 
relative to the target RPUPS for these groups (Figure C-12). Group 1 has the greatest dc\ iation 
above the target while Group 10 has the greatest deviation below the target (see the black bars in 
Figure C-12). The groups are presented in the same order for all subsequent figures. 

In Figure C-12, four groups (Groups 1 through 4) exceeded the unadjusted 1992 RPUPS 
target for total Medicare services, with Group 1 approximately 4 percent over the target. In most 
cases, adjusting for health status makes groups' performance appear worse. After adjusting for 
health status. Groups 6 and 8 switch from being below to being above their targets, while Group 
1 switches from being above to being below its target. Truncation of high cost outlier . 
reimbursements also leads to mixed results in the apparent performance of the groups. 

For physician and supplier services. Figure C-13 shows that eight of the ten groups have 
unadjusted 1992 RPUPS under the target, suggesting that groups may be able to better manage 
this component of services. In most cases, health status adjustment appears to worsen the 
performance. 

From 1992 to 1993, more groups were under their unadjusted RPUPS target for all 
services, with Group 10 over 10 percent below target (Figure C-14). After adjusting for health 
status, Groups 1 and 4 appear better relative to their targets. For the rest of the groups, health 
status adjustment leads the groups to appear worse in comparisons to targets. Overall, truncation 
tends to weakens apparent performance as well. 

Similar to the previous year, groups tend to be more successful in being under target for 
physician and supplier services (Figure C-15). Only Group 9 was above the unadjusted RPUPS 
target. However, as before, health status adjustment typically does not improve performance 
(with the exceptions of Groups 1 and 4). 

For 1991 to 1993, eight groups performed below the unadjusted 1993 RPUPS target for 
all services (Figure C-16). Group 10 appears to perform best, being almost 10 percent under 
target. Adjustments to target for health status and truncation have effects that are similar to those 
observed previously. For physician and supplier services (Figure C-17), eight groups perform 
below target for unadjusted RPUPS. Adjustment for health status improves the performance of 
Groups 1 and 4 but weakens the performance of the remaining groups. 

Overall, for these performance years. Groups 8 and 10 were most successftil in 
performing below the target for all services, while Group 2 was least successful. In terms of 
physician and supplier services. Groups 5 and 10 had the best performance, and Groups 2, 6, and 
9 the worst over the three years. 

If groups exceed their unadjusted target, one explanation may be that their case-mix has 
gotten worse. Similarly, if they are under their target, one explanation may be that their case-mix 
has gotten better. Adjustment for health status should help to eliminate deviance due to this 



c-11 



factor. We attempted to adjust for health status in order to see the extent to which apparent 
performance would change. In general, we found that group performance tended to worsen after 
health status adjustments. 

At this point we cannot be sure that the health status adjustment approach we explored is 
adequate. Some of our concern stems from the concentration of patients and reimbursements in 
so few of the health status cells, as was shown in Figures C-3 through C-5. This pattern was 
observed for all of the groups, and even the random beneficiary samples drawn from each market 
area. 

Other concerns relate to findings for the random samples in our study. For .any case-mix 
adjustment method, the health status in a random sample of beneficiaries should be expected to 
remain rather constant from one year to the next. In order to test this, we drew a random sample 
of beneficiaries from each of our groups' market areas, and applied the same health status 
adjustment methodology to them. Without exception, the apparent severity of illness in these 
random beneficiary samples increased over the three-year period. Severity of illness for the ten 
groups appeared to increase as well, but at a lower rate than that observed in their respective 
markets. 

There could be several explanations for this. One is that coding and documentation may 
have become more accurate and extensive over time. This would have consequences for the 
ACQ classificafions if the number of diagnosis codes presented on claims increased over this 
period. Indeed, more ICD-9 codes were reportable on facility claims as of 1992, and we used 
diagnostic codes from all places of service in the case-mix adjustments. Also, providers may be 
exhibiting greater sensitivity to accurate coding as a by-product of the implementation of MFS, 
or the fact that claims processing procedures of payers other than Medicare may have become 
more demanding. 

The faster rate of increase in the severity of illness for the market versus the ten groups 
may be due to the fact that the coding practices of providers in the market lagged behind that of 
the groups at the start of the three-year period. Subsequent improvements in the coding practices 
of the average provider may have had the effect of artificially lowering the observed performance 
of the groups. Since the majority of our groups have well-developed managerial and 
administrative support structures, they were likely coding more accurately and completely before 
the average provider in the market. As mentioned above, the implementation of MFS eliminated 
payment differentials based on local codes and physician specialties. This change might have 
affected the 10 groups and their markets differently, so that the severity of illness in the markets 
increased at a faster rate. 

A second explanation may be that ACG classification may be an inappropriate health 
status adjustment tool for Medicare beneficiaries. The ACG system was first developed for 
assessing the provision of ambulatory care, and was inifially based on using diagnoses recorded 
during ambulatory encounters. The approach used in this project included inpatient diagnoses as 



C-12 



well, and the Medicare population uses relatively more inpatient services. In addition, because 
most of the 10 groups are sophisticated tertiary/quaternary providers, the patients seen by these 
groups require even more inpatient services than Medicare beneficiaries in general. As a result, 
any weaknesses in using the ACG system in assessing the use of inpatient services is exacerbated 
for the groups at hand. 

We continue to explore the application of ACGs and other case-mix methodologies to 
GVPS. Other health status adjusters could ultimately prove more appropriate for the Medicare 
beneficiary population. 

F. Adjustments for Changes in Pricing 

Medicare make changes to both its procedure codings and their corresponding 
reimbursement values every year. Under GVPS the apparent stability of RPUPS could be 
affected by such changes. In addition, there are yearly changes in reimbursements across 
geographic areas, for example due to chances in wage scales. To deal fairly with these issues it 
may be necessary to control for geographic factors and for Medicare changes in coding and 
pricing. 

In order to standardize claim reimbursements in the light of geographic differences, we 
would need to divide our reimbursements by appropriate geographic factors, which vary by type 
of service: 

• Inpatient hospital payments are adjusted for each MSA or rural county in a state; 

• Physician payments are adjusted by carrier defined localities; 

Other services are cost-based but would require a similar adjustment. 

To control for changes in Medicare pricing and procedure codings which show up over 
time, a refined approach would be to create two RPUPS for each fiscal year. The one reflecting 
actual payments would be compared with the GVPS target to determine performance and 
whether or not a reward is to be granted. The second RPUPS would be created a year later 
reflecting simulated payments per claim after crosswalking to new procedure and pricing files to 
show what would have been paid if the new year's rules had been in effect. This latter RPUPS 
would form the baseline measure on which the next GVPS target would be based. 

The following example shows how this might work: 

Actual code: 1234 RVU = 1 .00 (CF=30) Reimbursement = $30.00 
Changed code: 1234.1 RVU =1.10 (CF=33) Reimbursement = $33.00. 



C-13 



In the example a reimbursement of $30.00 is used in calculating the RPUPS for the year of the 
claim. However, the next year the payment is counted as $33.00 in calculating the baseline 
RPUPS for computing GVPS target, this to reflect that the cost for the corresponding procedure 
during the new year would have been the higher amount. 

In our statistical analyses we compare both original and adjusted RPUPS to learn whether 
refinements to remove price effects actually improve stability, as we hypothesize that they will. 
Obviously, geographic adjustments would be unnecessary if wage increases and other regional 
factors changed more or less in unison. After all, national changes are suppose to be dealt with 
through the performance standards. Also, it must be admitted that a retrospective effort to adjust 
for pricing and procedure changes could be detrimental to the underlying goals of the GVPS 
program by diminishing evidence "procedure creep," whereby providers make upward 
modifications in their procedure codes to compensate for expected payment reductions due to 
refinements in Medicare's reimbursement systems. 

G. Comparisons of Procedure Mix: Local Averages Versus a Provider'^ Patients 

We have proposed models in which the rate of increase applied to a group's baseline 
value of RPUPS would be based on the average projected rate of increase for the local area. This 
is in keeping with the premise that groups could actively manage their patients' care, and two 
related assumptions: 

The group's management expertise includes primary care, or evaluation and management 
services; 

• The group's patients are broadly representative of the local Medicare population. 

The potential for these assumptions to be violated raises two types of questions regarding 
the GVPS models: 

Should criteria for eligibility include tests about a group's mix of patients or services, and 
the degree of similarity to market averages? 

How robust is the basic model in which market level increases are applied to specific 
groups' patients? 

Section IX of the report discusses possible eligibility criteria that would require expertise in the 
management of patients' services. The criteria could be expanded to make sure a group does 
resemble the market, in which case the average rate of increase would apply. However, some 
differences can be expected for many, if not all groups. We briefly examine two possible 
methods to adjust rates of increase for a group's patient population that does not resemble 
sufficiently the local patient population. One method is based on patient case mix, and a similar 
alternative is based on procedure mix. 



C-14 



The numerator of a group's RPUPS includes all reimbursements for services to the 
patients seen by the group, regardless of the provider. In the proposed model, a group's patient 
population is basically a small subset of the "local" Medicare population, since areas are defined 
by the patients' coumies of residence. Therefore, the expected rate of expenditure growth at the 
market level is an average that includes the provider's own patients and the array of services they 
use. 

The expected rate of growth in expenditures can differ significantly across categories of 
service and perhaps categories of patients. Reimbursements for services can be arrayed along 
dimensions that include type or place of provider, and type of procedure. For example, Medicare 
services can be divided into inpatient hospital, outpafient facility, home health, physician, etc. 
Furthermore, reimbursements within these groupings can be divided by procedure. For example, 
physician reimbursements can be divided into more refined categories. Patients can be arrayed 
along categories like those used in this study, e.g., reason for entitlement or diagnostic mix. 

Whichever i.iXay is chosen, i.e., services or patients, the reimbursements at the market 
level and the group level can be categorized accordingly. We can calculate the distribution of 
reimbursements across the categories. For example, at the market level, 60 percent of 
reimbursements may be for inpafient hosphal. For a particular group, 70 percent of the 
reimbursements may be for inpatient hospital. Assuming growth in expenditures for inpatient 
hospital is different than the average for other services, we can reweight the distribufion of 
reimbursements at the market level to mimic that of the group. In the example, we could 
increase the weight given to inpatient hospital reimbursements when setting the market level rate 
of increase that is applied to the group. 

A simpler adjustment could be based on Part A versus Part B reimbursements. More 
elaborate adjustments could be based on type of procedures, such as the method of service 
classification developed by Berenson and Eggers for HCFA. They divide physician procedures 
into: Evaluation and Management, Procedures, Imaging, Tests, and Other. Each of these 
categories also is divided into between four and eight subcategories. Alternatively, 
reimbursement rates could be weighted in an analogous fashion to match the distribution of 
patients seen by the group. 

The effect of refinements such as these would be to project the rate of increase for the 
group's particular mix of patients (measured as patient or service mix). The increases still would 
be based on the experience and expectations for the average provider in the market, adjusted to 
simulate a hypothetical average provider with the particular group's mix of patients. 



C-15 



TABLE C-1: AMBULATORY CARE GROUP (ACG) CATEGORIES 


AGO 


ACG DESCRIPTION 


1 


Acute Minor, Age < 1 


2 


Acute Minor, Age 2-5 


3 


Acute Minor, Age 6+ 


4 


Acute: Major 


5 


Likely to Recur, Without Allergies 


6 


Likely to Recur, With Allergies 


7 


Asthma 


8 


Chronic Medical, Unstable 


9 


Chronic Medical, Stable 


10 


Chronic Specialty 


11 


Ophthalmological/Dental 


12 


Chronic Specialty, Unstable 


13 


Psychosocial, Without Psychosocial Major 


14 


Psychosocial, With Psychosocial Major, Without Psychosocial Minor 


15 


Psychosocial, With Psychosocial Major, With Psychosocial Minor 


16 


Preventive/ Administrative 


17 


Pregnancy 


18 


Acute Minor and Acute Major 


19 


Acute Minor and Likely to Recur Discrete, Age < 1 


20 


Acute Minor and Likely to Recur Discrete, Age 2-5 


21 


Acute Minor and Likely to Recur Discrete, Age > 5, Without Allergy 


22 


Acute Minor and Likely to Recur Discrete, Age > 5, With Allergy 


23 


Acute Minor and Chronic Medical: Stable 


24 


Acute Minor and Eye/Dental 


25 


Acute Minor and Psychosocial Without Psychosocial Major 


26 


Acute Minor and Psychosocial With Psychosocial Major, Without Psychosocial Minor 



C-16 



ACG 


ACG DESCRIPTION 


27 


Acute Minor and Psychosocial with Psychosocial Major and Minor 


28 


Acute Major and Likely to Recur Discrete 


29 


Acute Minor/ Acute Major/Likely to Recur Discrete, Age < 2 


30 


Acute Minor/ Acute Major/Likely to Recur Discrete, Age 2-5 


31 


Acute Minor/Acute Major/Likely to Recur Discrete, Age 6-1 1 


32 


Acute Minor/ Acute Major/Likely to Recur Discrete, Age > 5, Without Allergy 




Acute Minor/Acute Major/Likely to Recur Discrete, Age > 5, With Allergy 


34 


Acute Minor/Likely to Recur Discrete/Eye & Dental , 


35 


Acute Minor/Likely to Recur Discrete/Psychosocial 


36 


Acute Minor/ Acute Major/Likely to Recur Discrete/Eye & Dental 


37 


Acute Minor/Acute Major/Likely to Recur Discrete/Psychosocial 


38 


2-3 Other ADG Combinations, Age < 17 


39 


2-3 Other ADG Combinations, Males Age 17-34 


40 


2-3 Other ADG Combinations, Females Age 17-34 


41 


2-3 Other ADG Combinations, Age > 34 


42 


4-5 Other ADG Combinations, Age < 17 


43 


4-5 Other ADG Combinations, Age 17-44 


44 


4-5 Other ADG Combinations, Age > 44 


45 


6-9 Other ADG Combinations, Age < 6 


46 


6-9 Other ADG Combinations, Age 6-16 


47 


6-9 Other ADG Combinations, Males Age 17-34 


48 


6-9 Other ADG Combinations, Females Age 17-34 


49 


6-9 Other ADG Combinations, Age > 34 


50 


10+ Other ADG Combinations 


51 


No Visits and/or No ADGs 



Source: Weiner, J.P. "A CHnician's Guide to the Johns Hopkins Ambulatory Care Group (ACG) 
Case-Mix System". 1992 Johns Hopkins University. 



C--I7 

Table C-2: Effects of Changes in Case Mix* on RPUPS Values (One Group) 



1991 Mean Dollars Per Cell 



Scope of Services 


1991 Case Mix 


1992 Case Mix 


1993 Case Mix 


Physician/Supplier 


$1,810 


$1,873 


$1,891 


% Change from 1991 




3.5 


4.5 


All Medicare 


$6,085 


$6,330 


$6,390 


% Change from 1991 




4.0 


5.0 



1992 Mean Dollars Per Cell 



Scope of Services 


1991 Case Mix 


1992 Case Mix 


1993 Case Mix 


Physician/Supplier 


$1,681 


$1,743 


$1,758 


% Change from 1991 




3.7 


4.6 


All Medicare 


$6,490 


$6,763 


$6,838 


% Change from 1991 




4.2 


5.4 



1993 Mean Dollars Per Cell 



Scope of Services 


1991 Case Mix 


1992 Case Mix 


1993 Case Mix 


Physician/Supplier 


$1,600 


$1,659 


$1,676 


% Change from 1991 




3.7 


4.8 


All Medicare 


$6,620 


$6,906 


$6,984 


% Change from 1991 




4.3 


5.5 



Source: National Claims History file, 1991-1993. 
* Derived from Ambulatory Care Groups. 



C--18 



Table C-3: Effects of Changes in Case Mix* RPUPS Values, With Minimum 

Threshold** 



1991 Mean Dollars Per Cell 



Scope of Services 


1991 Case Mix 


1992 Case Mix 


1993 Case Mix 


Physician/Supplier 


$2,276 


$2,325 


$2,395 


% Change from 1991 




2.2 


5.2 


1 992 Mean Dollars Per Cell 


Scope of Services 


1991 Case Mix 


1992 Case Mix 


1993 Case Mix 


Physician/Supplier 


$2,118 


$2,166 


$2,234 


% Change from 1991 




2.3 


5.5 


1 993 Mean Dollars Per Cell 


Scope of Services 


1991 Case Mix 


1992 Case Mix 


1993 Case Mix 


Physician/Supplier 


$2,029 


$2,075 


$2,141 


% Change from 1991 




2.3 


5.5 



Source: National Claims History File. 



* 

** 



Derived from Ambulatory Care Groups. 

Minimum reimbursement threshold to provider of $120. 



C-19 



Table C-4: Effects of Changes in Case Mix* RPUPS Values, With Outlier 

Tlireshold** 



1991 Mean Dollars Per Cell 



Scope of Services 


1991 Case Mix 


1992 Case Mix 


1993 Case Mix 


Physician/Supplier 


$1,709 


$1,762 


$1,780 


% Change from 1991 




3.1 


4.2 


All Medicare 


$5,245 


$5,425 


$5,476 _ 


% Change from 1991 




3.4 


4.4 



1992 Mean Dollars Per Cell 



Scope of Services 


1991 Case Mix 


1992 Case Mix 


1993 Case Mix 


Physician/Supplier 


$1,602 


$1,654 


$1,670 


% Change from 1991 




3.2 


4.2 


All Medicare 


$5,396 


$5,583 


$5,640 


% Change from 1991 




3.5 


4.5 



1993 Mean Dollars Per Cell 



Scope of Services 


1991 Case Mix 


1992 Case Mix 


1993 Case Mix 


Physician/Supplier 


$1,543 


$1,595 


$1,612 


% Change from 1991 




3.4 


4.5 


All Medicare 


$5,461 


$5,654 


$5,711 


% Change from 1991 




3.5 


4.6 



Source: National Claims History File. 

* Derived from Ambulatory Care Groups 

** Outlier thresholds are $10,000 for physician and supplier servies, and $30,000 for all 
services. 



Figure C-1: Percentage Decrease in Patients Seen by Omitting Patients with < $120 to Provider 

(Physician and Supplier Services) 

Provider 

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 



0% 



-10% 



-20% 



-30% 



-40% 



-50% 



-60% 



-70% 



-80% 



-90% 



-100% 



Source: National Claims History file, 1992. 



Figure C-2: Patient Capture Ratio, Omitting Patients With < $120 to Provider 

(Physician and Supplier Services) 



t— 1 1 1 1 1 1 r-1 1—1 1 1 1 1 1 " I " I " ! " " 'I i " " i " i " " I " " I " (" i U U i II 1 II I I II I' " M " " II i " " i M I M " I " ; " " I " I " " ■■ I " I " " I " i 

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 
Source: National Claims History file, 1992. Provider 



Figure C-3: Breakdown of One Group's Patients by Health Status 



0.35 



0.3 



0.25 



0.2 



0.15 



0.1 



0.05 



D = Disabled by ACG 
A = Aged by ACG 







I I l- M - l I r i- l I I I M 11 l -l I h i M I I I l-l I I l - l I I l - l -l l"! "! I I I I I 'l "!- ! -!"! hi W l "l I I'l l" l I I I l "l I I I' ' I-I I I I' 



JUL 



l"l l lli I |B| iDi. 



D 

C/) 
lU 



I 

Q 



00 

6 



CO 



O 

Q 



CM 



CD 
CM 



CM 
CO 

I 

a 



CD 
CO 



O 



CO 



in o 
< < 



CO 



T- CM 



in 



CD 
CM 



CO 
CO 



CO 



I 

< 



in 



cn 



Source: National Claims History file, 1992. 



CNI 

1 

a 



Figure C-4: Breakdown of Reimbursements by Patients' Health Status, for One Group 

(Physician and Supplier Services) 



< 
o 



c 
E 

0) 
CO 

t 

E 



c 
o 

o 

Q. 
O 

a. 



0.35 



0.3 



0.25 



0.2 



0.15 



0.1 



0.05 



D = Disabled by ACG 
A = Aged by ACG 



I I I I I I I 1 -1 I I I I M I 
Q ^ 00 CN 

Q 6 1 



1 1 1 1 1 1 1 1 I I 1 I I I l-l 1 1 I l -l-l I I l " l I 



I I I I l - l I I l" l I I l - l 'I I I I l-l I I I l - l I t M - l I I M I l-l' 



" i I I I I' ll 1 i " i I I I m 



CD O 



D 



CM 

I 

D 



00 
CM 



C-4 
CO 



CD 
CO 



O 



00 



I 

< 



in 

CM 



CVJ 



CO 
CO 



CO 



I 

< 



If) 



cn 



Source: National Claims History file, 1992, 



I 

U 



Figure C-5: Breakdown of Reimbursements by Patients' Health Status for One Group 

(All Medicare Services) 



0.35 



0.3 



i 0.25 



0.2 



0.15 



0.1 



0.05 



D = Disabled by ACG 
A = Aged by ACG 







I II I ll - I ll l -l-l I I I l -l I I l - l I l - l I I I I I I I I I M I I l- l - l l -l l" l I I I I I 



Q 



D 



00 

I 

Q 



Q 



CD 
t— 

Q 



O 

I 

Q 



CO 
CN 



CO 



CO 



O 



03 



in 

t 

< 



I 

< 



CO 



■I l° l l°l 



■ i ni n -i 1-1 j P i I I |n| I |-|" m I I l " l I l"l I I I I"'"' 



CM 

I 

< 



in 
eg 



CM 



CO 
CO 



CO 



< 



in 



CD 



Source: National Claims History file, 1992, 



Figure C-6: Deviations from Targets for 78 Providers in 10 Market Areas: 1991 to 1992. Excluding 

Patients with <$1 20 to Provider 

-QS- 



C.3 



0.5 



0.4 -- 



0.3 



0.2 - 



0.1 - 



♦ ♦. 



0^ 



-0.2 



-0.1 



♦ 

♦ 



♦ ♦ ♦ 



♦ ♦ 



♦ ♦ 



♦ 







♦ 
♦ 



♦ 



-0.1' 



-0.2 



0.1 



0.3 



OtS- 

Percentage Deviations from Targets for Physician/Supplier Reimbursements 



Source: NCH file, 1991 and 1992. 



CM 
I 

U 



i2 
c 
a> 
E 

!2 

3 
X) 

E 
'S 
q: 

s 

o 



tn 

q3 

E 

(0 

c 
g 

(0 

> 

0) 
Q 

Q) 
O) 

to 
c 

Q) 
O 
u. 
(U 
Q. 



-0 



25 



Figure C-7: Deviations from Targets for 78 Providers in 10 IVIarket Areas: 1992 to 1993, Excluding 

Patients v^ith < $1 20 to Provider 



-0t2- 



0.1 



♦ 
♦ 



-0.2 



-0.15 



-0.1 ♦-o.Qe 

♦ t * 

♦ ♦ ♦ - ♦ 

► ♦ 



♦ ♦ 



-0.1 



-0.2 



-0.3 



-OA- 



-♦I — 

0.05 



0.1 



0.15 



Percentage Deviations from Targets for Physician/Supplier Reimbursements 

Source: NCH file, 1992 and 1993. 



I 

KJ 



(D 

E 
a> 

i2 

3 
X> 

E 

B 
a> 

E ■ 

m 

■c 
o 

ia 

*l 
Q 

0) 

D) 

TO 
' 

C 

<u 
at 

Q. 



25 



Figure C-8: Deviations from Tarc ets for 10 Selected Providers: 1992 to 1993, Excluding Patients 



-0.2 



K 



with <$120 to Provider 



-0.15 



-Or4- 



0.05 



R 



-0.1 



-0.05 



M 



-0.05 



-0.1. 



-0.15 



-0.2 



♦Q 



N 



0.05 



^ Percentage Deviations from Targets for Physician/Supplier Reimbursements — ^ 

Source: NCH file. 1992 and 1993. 



00 
I 

U 



Si 

§ 
E 

S2 

3 
J3 

E 
■<u 
a: 

I 

J2 -0 

Q) 

n 
E 

M 
C 

o 

•3 
CO 

'I 

Q 

0) 
O) 

iS 
c 
a> 

B 
a> 
a. 



14 



Figure C-9: Deviations from Targets for 10 Selected Providers: 1992 to 1993. Excluding Patients 

with < $120 to Provider 



-0.12 



K 



-0.1 



-Or-1- 



-0.08 



-0.06 



0.05 -- 



M 



-0.04»j -0.02 
♦R 

Q 



-0.05 



-0.1 



-0.4^ 



Percentage Deviations from Targets for Piiysician/Supplier Reimbursements 

Source: NCH file. 1992 and 1993. 



002 
N 



Figure C-10: Deviations from Targets for 10 Selected Providers: 1991 to 1993, Excluding Patients 

with < $120 to Provider 



c 

Q) 

E 

0) 

w 
E 

0) 

al 

TO 
O 



0) 

cn 

TO 



E 

to 
c 
g 

■> 

(U 
Q 
<u 

O) 

jg 
c 

Q. 



0.4 

0.3 - 

0.2 - 
♦ 

0.1 



-0.3 



-0.2 ♦ -^.tV^^^^IW 

♦ ♦ «♦ 

♦ ♦ . ♦0.1^«- 



♦ ♦ 



♦ ♦ 



♦ 
♦ 



♦ 



♦ 
♦ 



# ♦ 



0.1 



♦ 0.2 



0.3 



0.4 



♦ ♦ 



-0.2 
-0.3 ^ 
-0.4 



Percentage Deviations from Targets for Physician/Supplier Reimbursements 

Source: NCH file, 1991 and 1993. 



o 

I 

U 



Figure C-11: Deviations from Targets for 10 Selected Providers: 1991 to 1993, Excluding Patients 

with <$1 20 to Provider 



J2 
c 
a 

E 
a 

XX 

E 
« 
a: 

3 -0 



a 
ra 

E 
o 
4: 
w 
c 
,o 

re 
■> 

0) 

Q 

0) 
D) 

ra 
<*-* 
c 

0) 

u 

0) 
Q. 



K 



— I — 
-0.25 



-0.2 



-0.15 



-0.1 



I ♦ 



M 



-^1-5- 



0.1 



0.05 



-0.05 Q 



-0.05 



-0.1 -- 



-0.15 -- 



-0.2 



-0.25 



-0.3 



-0t36- 



Percentage Deviations from Targets for Physician/Supplier Reimbursements 



0.05 



N 



Source: NCH file. 1991 and 1993. 




Source: NCH file, 1991 and 1992. 



Figure C-13: Deviations from Targets for Physician/Supplier Reimbursements for 10 Providers in 10 

Market Areas: 1991 to 1992 



0.06 



0.04 



0.02 



-0.02 - 



-0.04 




-0.06 



-0.08 

Source: NCH file, 1991 and 1992 



Unadjusted 
HS Adjusted 



Figure C-14: Deviations from Targets for Total Reimbursements for 10 Market Areas: 1992 to 



0.08 - 




-0.12 J 

Source: NCH file, 1992 and 1993. 



Figure C-15: Deviations from Targets for Physician/Supplier Reimbursements for 10 Providers in 10- 
<r Marl<et Areas: 1992 to 1993 

CO . . 

I 




-0.08 



Source: NCH file, 1992 and 1993 



0.15 



Figure C-16: Deviations fron Targets for Total Reimbursements for 10 Providers in 10 Market 

Areas: 

1991 to 1993 : 





I Unadjusted 
IHS Adjusted 
□ HS/ Outlier 



Source: NCH.file, 1991 and 1993. 




Source: NHCfile: 1991 and 1993. 



ens LIBRSRV 




III IIIIIIIIIIIlL, 

3 aOTS QQQ1E7E1 3